Atmospheric Conditions Leading to an Exceptional Fatal Flash Flood in the Negev Desert, Israel

The study deals with an intense rainstorm that hit the Middle East between 24 and 27 April 2018. The storm reached its peak over Israel in April 26, when it produced a heavy flash flood that took the lives of ten people. The rainfall observed in 10 the southern Negev was comparable to the long-term annual rainfall there. The timing of the storm is also unique, at the end of the rainy season, when rain is relatively rare and spotty. The study analyses the dynamic and thermodynamic conditions that made this rainstorm one of the latest spring severe events in the region during the last 3 decades. The synoptic background was an upper-level cut-off low that entered the region from west, along 30oN latitude, which is rather exceptional for such systems in the late spring. While approaching the Levant, it slowed its movement from ~10 to <5 ms-1. 15 On the day of maximum intensity, the radius of the cyclone shrank to 275 km. The effect of the small radius was estimated by the measure of its curvature vorticity (MCV), which was the largest among the spring rainstorms during the latest 33 years. The lower levels were dominated by a north-westerly wind that advected moist air from the Mediterranean inland. During the approach of the storm, the atmosphere over Israel became unstable, with instability indices reaching values favourable for thunderstorms (CAPE = 909 J Kg-1, LI = -4.9 K, SI = -2.7 K and MKI = 30 K), and the precipitable water increased from 17 20 to 30 mm. The latter is explained here by a combined effect of the lower-level moisture advection and a mid-level band of tropical moisture that entered Israel above it. Three major rain centres were active during April 26, two of them were non-orographic, which is unusual for this type of system. This is explained by the dominance of sub-synoptic features, found in the 0.25 resolution data of ERA5 that were used to derive Omega and MKI maps. The build-up of static instability is explained by a -5 K temperature anomaly over the region, 25 caused by a northerly flow east of the blocking high over North Europe that transported cold air over the Mediterranean water. The unique intensity of this storm is attributed to an amplification of a mid-latitude disturbance, which produced a cut-off low, with its implied high relative vorticity, low upper-level temperatures and slow progression. All these, combined with the contribution of several moisture sources, led to extreme dynamic and thermodynamic conditions favourable for this exceptionally severe rainstorm. 30 https://doi.org/10.5194/nhess-2020-94 Preprint. Discussion started: 29 April 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
On April 26, 2018 an extremely intense rainstorm hit Israel. Ten people were killed by a raging flash flood in Tzafit creek, in the northeast of the Negev Desert (denoted in Fig. 1), that resulted from this storm. The rainfall depth during this storm exceeded 20 mm over ~ half of Israel, over most of which the monthly long-term mean does not exceed 10 mm.
Flash flood forecast and warning is a challenging task, in particular over remote arid areas, where meteorological radar 35 coverage is sparse. Though flash flood warning has much improved in the last few decades, it is still the most weather-related fatal hazard globally (Montz et al. 2002), with 1,075 fatalities between 1996 and 2014 in the United States alone (Terti et al 2019). A significant part of flash flood victims is caused by incorrect assessment or misjudging of the risks of these rapidly evolving phenomena (Sharif et al. 2015;Becker et al. 2015;Diakakis et al. 2018).
The majority of the annual precipitation in Israel is associated with Mediterranean cyclones, while reaching its eastern part 40 (i.e., Cyprus Lows, HMSO 1962;Saaroni et al. 2010;Zappa et al. 2015), two-thirds of which occur during December through February (Alpert et al. 2004;Ziv et al. 2006). The rest of the annual rainfall occurs in the transitional seasons and is contributed by precipitating tropical synoptic-scale systems and by Cyprus Lows. A significant part of them occur in the desert areas and are characterized as intensive rain events of small spatial extent and short duration, some of which produce flash floods (Kahana et al. 2002;Morin 2006, Greenbaum et al. 2010). 45 The aim of this study is to analyse the dynamic and thermodynamic conditions that turned this rainstorm into one of the latest spring severe events in the region during the last 33 years, and to explain its unique features, particularly the rain spatial distribution. The study concentrates on the Negev Desert, which is the southern half of Israel, located to the southeast of the Mediterranean.
The article is organized as follows: in the next section, we present the data and methods used to analyse the event. In section 50 3 we describe the event and identify the unique dynamic and thermodynamic conditions that lead to the severe convection, as well as the sources of moisture for the rain formation in this storm. Discussion and conclusions are presented in section 4.

Materials
The study area is the Negev Desert, covering the arid part of Israel, south of 31.2° N, and the Judean Desert (the latter is denoted in Fig. 1), over which floods took place. The synoptic scale data, including satellite imagery, cover the entire 55 Mediterranean Basin and South Europe.
The study covers a storm that influenced the study region between 24 and 27 April 2018, in particular to 26 April, when it reached its maximum intensity. This storm is compared to other 11 storms, spread over 28 days, in which discharge was observed over the study area, during April and May for the years 1986 -2018, entitled hereafter the 'reference period'.
The rainfall analysis is based on the rain measuring network of the Israeli Meteorological Service (IMS Fig. 1) and the rain 60 maps are taken from the IMS publication (Porat et al. 2018). The radar data is taken from the IMS radar, in terms of rain intensity (Marra and Morin 2015). https://doi.org/10.5194/nhess-2020-94 Preprint. Discussion started: 29 April 2020 c Author(s) 2020. CC BY 4.0 License.

65
The data of floods during the reference period is based on the 11 maximum discharge storms measured over the Negev Desert, the Dead Sea valley and the Arava by hydrometric stations of the Israeli Hydrological Service for April-May of the period 1986-2018.
The atmospheric processes responsible for flash flood-producing rainstorms act in concert at the synoptic and sub-synoptic scales. The data for the synoptic maps, including sea level pressure (SLP) and upper-level fields, geopotential height (gph) and 70 vertical velocity (Omega), were taken from NCEP/NCAR reanalysis 1 database, with 2.5° x 2.5° spatial resolution (Kalnay et al. 1996;Kistler et al. 2001). The analysis of the mesoscale features is based on ERA5 ECMWF reanalysis data base (C3S, 2017, https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels), with 0.25° x 0.25° resolution, including wind field, relative vorticity, Omega and specific and relative humidity.
Several methodologies have been offered to quantify cyclone intensity and activity, Raible et al. (2007) for example, used the 75 mean gradient between the cyclone core and a radius of 1,000 km as a measure of cyclone intensity. Here the dynamic factors Ju d e a n M t.
where xc is curvature vorticity and V is the tangential wind speed. The cyclone radius, R, is that of the outer-most closed 80 isohypse (when derived with 15 m intervals). The tangential wind speed was calculated using the geostrophic relation, i.e., where g and f are gravitational acceleration and Coriolis parameter, respectively, and D is the depth of the cyclone. Inserting V from Eq. (2) in Eq. (1) yields: http://weather.uwyo.edu/upperair/sounding.html. These are: • Lifted Index (LI): Temperature difference between the environment and an air parcel lifted adiabatically at a given pressure height in the troposphere (lowest layer where most weather occurs) of the atmosphere, usually 500 hPa (Galway 1956). 90 Negative values indicate instability.
• The Showalter stability index (SI) is similar to the lifted index (LI), but while the LI starts with the mean of the lowest 100-hPa AGL layer, the SI uses a parcel lifted from 850 hPa to 500 hPa. At 500 hPa the parcel temperature is subtracted from the sounding temperature. More negative SSI values indicate greater instability (Showalter 1953).
• Convective Available Potential Energy (CAPE): The integrated energy excess of an air parcel lifted adiabatically with respect 95 to its environmental temperature profile. (Moncrieff and Miller 1976). Values in the order of hundreds of J Kg -1 and more correspond to severe weather.
Maps of the Modified K-Index (MKI) were derived from the ERA5 ECMWF data, with a 0.25° resolution. The original K-Index (KI; Geer 1996), which combines instability and moisture availability, is used to predict severe thunderstorms in the where T and Td are temperature and dew point, respectively, RH is relative humidity and the subscripts refer to the respective pressure level (hPa). The modification gives more weight to the relative humidity at the 850 and 700 hPa levels.

Analysis 110
Precipitation extremes and strong winds are usually induced by large-scale atmospheric circulation, namely low-pressure systems with large pressure gradients (Jacobeit et al. 2017;Knippertz 2018) therefore, this study was initiated by analyzing the synoptic circulation system with which this storm was associated. During 25-27 April 2018 the lower-levels were dominated by a cyclone, centred east of Israel (see Fig. 2). The implied winds were north-westerly, hence advecting moist air from the Mediterranean inland. This flow was perpendicular to the topographic upslope of the Judean Mountains and the 115 northern Negev (see the water divide line denoted in Fig. 1), which activated convection within air parcels ascending there.
The rain in Israel started at April 25, reached its maximum intensity at April 26 and stopped at April 27 (IMS 2018). The spatial distribution of the rain deviated from the expected one in the sense that a considerable part of the major precipitative elements were observed over the downslopes of the mountain regions ( Fig. 1). The maximum rainfall was obtained in the rainshadow of the Negev Mountains, and the second most intense one was found in the Jordan valley, again, in the lee side of the 120 Samarian Mountains. The dominance of convective over orographic elements suggests that sub-synoptic scale factors took place in this storm. This subject is elaborated below.
Two complementing factors contributed to the rain formation. One is dynamic, i.e., vertical ascent, associated with the cyclonic system described below (Sec. 3.2). The other is thermodynamic, which is composed of instability and moisture supply, described in Sec. 3.3. 125

Synoptic evolution
Prior to the storm, on April 24, 2018, the lower levels of the atmosphere were dominated by a Red Sea Trough (RST). The RST is a low-pressure system extending from south toward the Eastern Mediterranean (EM) and the Levant (Ashbel 1938;Kahana et al. 2002;Tsvieli and Zangvil 2005). The trough axis was located east of the Levant region, enclosing a shallow cyclone, with a central pressure of 1008 hPa, over north western Saudi Arabia (see Fig. 2a). This surface system transported 130 warm and dry air from the east over the Levant. At the 500 hPa level, a closed cyclone was approaching the region from the central Mediterranean and slowed its progress during its arrival to the Egyptian coast, from ~9 ms -1 to ~4 ms -1 (see Fig. 3).
There are difficulties in reaching a common view on the climatology of extra-tropical cyclone activities, e.g., size, lifetime and speed (Lionello et al. 2002;Ulbrich et al. 2009), which stem from different identification and tracking methods. However, it should be noted that it is common to consider the average translation speed of extra-tropical cyclones is ~10 ms -1 (Laurencin 135 and Misra 2019). The corresponding 300 hPa potential vorticity (PV) field shows a pronounced positive anomaly, with a maximum of 1.13 PVU 140 ( Fig. 4). Following Campa and Wernli (2012), this feature typifies mid-latitude cyclones. This is rather exceptional if one takes into account the timing of the late spring and the latitude along which the system moved (~30° N).
Upon initiation of the storm, on April 25, the 500 hPa cyclone continued to slow its eastward motion, with an average speed of 3.5 m sec -1 and reached the Sinai Peninsula (Fig. 3). Concurrently, the surface cyclone that was imbedded in the RST deepened ( Fig. 2b) and moved northward. In the following day, on April 26, the storm attained its maximum intensity during 145 noontime. The upper air cyclone in that day continued its slow eastward progression, which created a north-westerly flow across the entire atmospheric column over the EM (Fig. 2c). The weakening of the horizontal pressure gradient at the surface, in tandem with a flattening of the 500 hPa cyclone, on the following day (April 27, Fig. 2d  The identity of the upper-level cyclonic system as a cut-off low can be deduced from the isolated PV maximum (Hoskins et 160 al. 1985), seen in April 26 00 UTC, 300 hPa map (Fig. 4). The slowing down of the upper-level cyclone while approaching the Egyptian coast is dynamically explainable in terms of point vortex interaction. This approach assumes that each point vortex induces a circular flow pattern that extends out of its core (e.g., Aref 1983). An example of interaction between twopoint vortices is in the case of a dipole which consists of a blocking high located north of a cut-off low (Yamazaki and Itoh 2013). The interaction within such a dipole (presented schematically in Fig. 5) is expressed by easterly flow, induced by the 165 blocking high, at the location of the cut-off low, and by easterly flow, induced by the cut-off low, over the blocking high. Between April 24 and 25, a large anticyclone that covered Europe, split into two separated anticyclones (Figs. 6a and 6b). The 170 eastern one, together with the cyclone over the south-eastern Mediterranean, formed a dipole-type structure system. The induced flow of each of the eastern anticyclonic and the cyclonic vortices slowed down the eastward propagation of each other against the westerly background flow. It should be noted that the deceleration of a cyclonic system is expected to further enhance its induced dynamic effects, i.e., divergence and vertical ascent (Palmen and Newton 1969;Ziv and Paldor 1999).    Figure 7 demonstrates the association between the dynamic factor (and the thermodynamic factor, as described in Sec. 3.3), represented by the 700 hPa Omega, at a meso-scale resolution of 0.25 , and the distribution of the rain systems, as detected 185 by the radar. In the morning of April 26, the most prominent feature is a zonally oriented cloud strip, extending over western Syria and north Israel (Figs. 7e and 7f). This strip was co-located with a region of negative Omega (ascendance of >10 Pa s -1 , Figs. 7a and b), implying that this rain system is supported by the dynamic factor.

Dynamic conditions
Toward noon time, this cloud system rotated cyclonically and became meridionally oriented, as can be inferred from both the radar imagery and the ascending area (Figs. 7g and 7c, respectively), which was parallel to the mountain ridges of Israel (Fig.  190 1). Three rain centers are noted within this system (Fig. 7g): one, the northernmost, was located at the Jordan valley, the central, spread along the mountains' upslope, and hence can be considered as forced by orography, and the southernmost, which caused the fatal flood, over a relatively flat terrain (denoted Tzafit in Fig. 1). It should be noted that the three rain centres are located within a region of negative Omega (ascendance, Fig. 7c), with an extremum value of ~ -10 Pa s -1 . This implies that these rain systems are dynamically supported. In the late afternoon, southern Israel was still under precipitative cloudiness (Fig. 7h), but 195 the dynamic signature, i.e., the vertical ascent dropped to ~ -0.1 Pa s -1 there (Fig. 7d).
The characteristics of the upper-level cyclone in this event were extreme for the pertinent season. An objective 500 hPa cyclone climatology for the Mediterranean region for 45 years by Campins et al., (2011) points at cyclone radii between 350 and 700 km. The radii of the 500 hPa cyclones for the 28 days in the reference period (defined in Sec. 2) vary between 275 and 1170 Km, with an average of 625 km (Table 1). On the day of the flood (April 26), the radius of the upper-level cyclone was the 200 smallest, 275 km, while located over Jordan (Fig. 3c). The effect of the small radius of the upper-level cyclone can be estimated by a measure of its curvature vorticity (MCV). A comparison of all the 28 days (Table 1) indicates that for the storm analysed here the MCV is the highest of them all.

Thermo-dynamic conditions
During the approach of the upper-level cyclone, the thermodynamic state of the atmosphere over the study region became 210 more and more favourable for convection. Figure 3 shows the values of leading instability indices, specified in Sec. 2, derived from Beit Dagan soundings, in 24-hour increments. They reflect a gradual increase toward April 25 12 UTC, while the upperlevel cyclone approached the Sinai Peninsula. At that time, the values of CAPE (779 J Kg -1 ), LI (-4.9 K) and SI (-2.7 K) indicate potential for thunderstorms. At the same time, the precipitable water over the study area, increased from 17 mm, while the upper-level cyclone approached the western tip of the Nile delta, to 30 mm in April 25 12 UTC. During April 26, the 215 moisture content remained between 26 and 27 mm, while the CAPE rose to 909 J Kg -1 . The sharp increase in the precipitable water can be explained by the wind shift to the northwest direction in the lower levels, which induced onshore flow throughout the troposphere (Fig. 2). The MKI distribution over the study region for the April 26, at 03, 09, 15 and 21 UTC, is shown in Figs. 7i-l, respectively. Values exceeding 25 o C, indicating potential for thunderstorms, are co-located with the major rain centres at the hours 09, 15 and 21 UTC. 220 The moisture transport is noted in the satellite imagery of the water vapor channel for April 25 21 UTC (Fig. 8). The major 225 feature is a moisture strip that extended from tropical latitudes to Saudi Arabia, curved cyclonically through Iraq, where it interacted with deep moist convection, and entered northern Israel from the east in the morning of April 26 (Fig. 7e). The moisture of this strip seems to be the core of the cloud systems that later on, entered central and south Israel through the eastern Mediterranean (Fig. 7f), including the rain system that reached Tzafit at the noon hours (Fig. 7g). The trajectory arriving at Tzafit at 1400 m ( Fig. 9) was located over this convective region 18-24 hours prior the flood. The passage of the air over the 230 Mediterranean Sea contributed further moisture to this rain system via the north-westerly winds at the lower levels, as can be inferred from the two lower trajectories shown in Fig. 9. Three moisture sources were responsible for the effectiveness of this rainstorm. One is of tropical nature, originated from East Africa, which were transported at the upper levels. The second, originated from the Fertile Crescent (western Iraq), via deep moist convection and the third is at the lower levels, from the Mediterranean Sea by the north westerlies. The combined moisture, in addition to its direct effect, enhanced the instability 235 described above.

Summary and Discussion 240
An intense rainstorm hit the Middle East between 24 and 27 April 2018, and reached its maximum intensity on April 26, when it produced a heavy flash flood that took the lives of ten people. The study concentrates on the Negev and the Judean deserts, Israel. The rainfall observed in the southern Negev was comparable to the long-term annual rainfall (<100 mm, IMS) there.
The timing of the storm is also unique, being at the end of the rainy season, when rain is relatively rare and spotty (Dayan and Sharon, 1980). The location of the surface cyclone was similar to the 'Syrian low', defined by Kahana et al. (2002) as one of 245 the major systems causing floods in the Negev Desert.
The study analyses the dynamic and thermodynamic conditions that made this rainstorm one of the latest spring severe events in the region during the last 3 decades. The dynamic factors addressed to quantify the severity of the event are vertical velocity and curvature vorticity. The stability was estimated by (a) standard indices (LI, SI and CAPE), derived from sounding in middle Israel and (b) the Modified K index (MKI, Harats et al 2010;Ziv et al 2016) that was mapped over the region. The 250 moisture was estimated by the vertically integrated precipitable water.
Three major rain centres were active during April 26 (Fig. 7g). One developed upslope the mountains, i.e., orographic, another in the Jordan valley and the third, which caused the fatal flood, over a relatively flat terrain. The distribution of the precipitating systems implies that synoptic and sub-synoptic atmospheric processes acted in concert. To resolve these scales, atmospheric data of various resolutions were used. The mesoscale distribution of Omega and MKI confirms that the three precipitating 255 systems were co-located with upward motion, and a maximum in the MKI.
The dynamic background of the storm was an upper-level cut-off low, with a PV maximum >1.1 PVU in the 300 hPa level (Fig. 4), typifying mid-latitude cyclones (Campa and Wernli 2012). This cyclone entered the Levant from the west along the North African coast, which is rather exceptional for late spring and the latitude along which the system moved. While approaching the Levant, this upper-level cyclone slowed its movement from typical extra-tropical cyclones of ~10 ms -1 260 (Laurencin and Misra 2019) to less than 5 ms -1 . On the day of the flood (April 26), the radius of this cyclone was the smallest, 275 km. The effect of the small radius of this cyclone was estimated by the measure of its curvature vorticity (MCV). A comparison of all 28 days in which floods were observed over the study area during the latest 33 years (Table 1) indicates that during this storm the radius was the smallest and the MCV was the highest of them all.
The lower levels were dominated by a cyclone, centred east of Israel (Fig. 2). The implied winds were north-westerly, hence 265 advecting moist air from the Mediterranean inland. During the approach of the upper-level cyclone, the atmosphere over the study region became favourable for convection. The instability indices reached values indicating potential for thunderstorms (CAPE = 909 J Kg -1 , LI = -4.9 K, SI = -2.7 K and MKI = 30 K). At the same time, the precipitable water over the study area increased from 17 to 30 mm.
In addition to the lower-level moisture advection, a mid-level band of tropical moisture, extending through Saudi Arabia, 270 curved cyclonically through Iraq, gained further moisture through deep convection, and entered north Israel in the morning of April 26 from the east. At noon hours this moist air band turned left (south and then southeast) over the lower-level Mediterranean moist air. The combination of both moisture contributions can explain the high precipitable water observed over the study region. The severe rain showers over Israel are attributed directly to the moisture contribution, and indirectly, to the activation of the pre-existing conditional instability. 275 The severity of this rainstorm can be explained by several factors: • The small radius of the upper-level cyclone, and its resulting large curvature vorticity. The latter, approximated by the MCV, was the largest among the 33-year data record (Table 1).
• The latitude along which the upper-level cyclone moved, i.e., 30 N. This is rather south of the common track of upper-level cut-off lows that affect the Mediterranean (e.g., Ziv et al. 2010). This reflects the extremely high amplitude of the upper-level 280 associated disturbance.
• The slow movement of the upper-level cyclone while crossing the region. Slow propagation of upper-level disturbances is expected to further enhance their induced dynamic effects, i.e., divergence and vertical ascent (Palmen and Newton 1969;Ziv and Paldor 1999).
• High static instability. The CAPE values reached 1000 J K g -1 . The origin of the instability is discussed below. 285 • Combined contribution of several moisture sources. An airstream of tropical origin was enriched by moist convection. This air merged with underneath Mediterranean moist air mass. The role that tropical moisture plays in Mediterranean rain systems has been raised by Krichak et al. (2007). A contribution of moisture from tropical latitudes was shown for a rainstorm over Israel in 6 April 2006 (Morin et al. 2007 , Fig. 3).
• The involvement of Sub-synoptic features. On top of the synoptic-scale factors, meso-scale features were found to activate 290 the main rain producing systems. The dominant lower-level cyclone during the storm resembles that of the 'deep cyclone to the east', which is one of the types of Cyprus lows defined by Alpert et al. (2004). Saaroni et al. (2010) showed that most of the rain associated with this type of low is obtained over the western slopes of the Judean Mountains. The existence of two out of the three non-orographic rain-producing systems emphasizes the sub-synoptic contribution.
An interesting point relates to the build-up of static instability under the influence of low to the east in the spring. The lower-295 level conditions were similar to the rainy phase of a Cyprus low. In the winter, cold air from south Europe interacts with the underneath warmer Mediterranean water and becomes unstable before entering Israel (Shay-El and Alpert 1991). In the late spring, Europe becomes warmer and the Mediterranean remains cool, so that instability is unexpected under the above conditions. In the case analysed here, the instability can be attributed to a -5 K temperature anomaly over the region, caused by a northerly flow east of the blocking high that transported cold air from North Europe (see Fig. 6, above). This 300 negative anomaly enabled the build-up of the observed instability.

Conclusions
The unique intensity of this storm can be attributed to an amplification of a mid-latitude disturbance, which produced a cutoff low, with high relative vorticity, low upper-level temperatures and slow progression. All these, combined with the contribution of several moisture sources, led to extreme dynamic and thermodynamic conditions favourable for this 305 exceptionally severe rainstorm. The severity of this rainstorm, and the casualties it caused, together with previous studies of similar rainstorms (some are quoted above), raise several conclusions, in particular for the EM region: • The late spring, though denoting the end of the rainy season, still bears the potential of mid-latitude synoptic-scale intense disturbances, which can produce heavy rainstorms. This is especially true whenever such an upper disturbance turns into a cutoff low. 310 • The combination of a cut-off low with small radius and large hypsometric depth, implies high curvature relative vorticity, with strong dynamical forcing on rain formation. • Quasi-stationary upper level systems allow moisture accumulation causing the increase in precipitation amounts from one day to the next.
• Moisture originating from tropical sources during such rainstorms enriches the mid-atmospheric levels, which makes the 315 rain formation less sensitive to availability of moisture in lower levels. Hence rain cells are expected not only over mountain upslopes, but also over low terrains such as the one that caused the deadly flood in Tzafit creek.
• The analyses of meteorological fields, in particular those attributed to meso-scale phenomena, are crucial to identify centres where torrential rains are expected. This makes regional models a central tool for short term prediction of floods and flash floods. 320 • Despite the universality of stability indices developed to illustrate the potential for convection, few of them require adjustments and modifications to fit the area being analysed. In this study, the modified KI version adopted for the eastern Mediterranean region, has shown to be a reliable predictor for convective rain centres and therefore a good precursor for floods.