Factors affecting isotopic composition of the rainwater in the Negev Desert, Israel

Authors


Abstract

[1] The study examines the δ18O of rain observed in the middle of the Negev Desert, Israel, based on seven rainstorms, associated with convective clouds resulting from midlatitude cyclones found over Trans-Jordan. The analysis is based on synoptic maps, vertical atmospheric cross sections, and air back trajectories. The rainwater sampling was done in temporal scale of hours. Four known factors were addressed here: the temperature effect, the amount effect, and two geographical effects: the marine and continental effects. The temperature effect is expressed by a positive relation between the lower level temperature and the isotopic composition. The amount effect, related to rainfall, is negative. The marine effect, expressed by a relation between the time the rain-producing air spends above the Mediterranean Sea, is positive and the continental effect, associated with the passage over land is negative. The linear relationships found in this study met the expectations, although only the amount effect was statistically significant. A fifth effect, entitled cloud top effect proposed here, relates to the height of the convective cloud layer top. It assumes that lower top prevents entrainment of upper cold air into the clouds, so the isotopic composition is enriched. The cloud top effect was found most statistically significant.

1 Introduction

[2] Stable isotopes within rainwater are used as tracers in atmospheric water to improve our understanding of the processes involved in moisture transport and rain production. There are a large number of observations of stable isotopes in precipitation with monthly time resolution all over the globe [e.g., Schootterer et al., 1996]. Others like, Araguas-Araguas et al. [2000], Gat [1996], and Lawrence et al. [2004], dealt with stable isotopes in atmospheric water vapor.

[3] It is well known that temperature is the main factor determining the isotopic composition; i.e., the higher the temperature in the moisture source, the heavier (enriched) the isotopic composition. The temperature depends on altitude, latitude, seasonality, and the geographic factors (mainly the identity of the source region as sea or land) [e.g., Clark and Fritz, 1999]. The path of air parcels from the source region on their way to the target area also influences the isotopic composition. An example for such an influence is the continental effect, expressed in a depletion in the isotopic composition of the cloud water (and the resulting rain), while a maritime air mass moves inland [Clark and Fritz, 1999].

[4] Another effect is the amount effect; i.e., the larger the rain amount, the more depleted the isotopic composition becomes. This effect was first reported by Dansgaard [1964], who attributed it to evaporation from the raindrops to the surrounding air because the lighter water molecules evaporate more easily. He proposed that in the beginning of a rain, the surrounding air is still dry so that evaporation is relatively intense, and subtracts large part of the lighter isotopes. After considerable rain, the air becomes almost saturated with water vapor so that the evaporation effect becomes much smaller. Later, others like Clark and Fritz [1999], showed the amount effect by comparing the isotopic composition of rain below and above 20 mm in a dry climate in Bahrain. Lee and Fung, 2007 explained the amount effect by the interaction between raindrop size, rain intensity, and water vapor near the planetary boundary layer (PBL). Dody et al. [2010] analyzed the relation between the isotopic composition (of the database used in the present study) and the instantaneous rain intensity and found no correlation between them. Miyake et al. [1968] reported that the exchange between raindrops and water vapor is more effective than the evaporation in enriching the isotopic composition.

[5] Previous studies, e.g., Cappa et al. [2003] and Gat et al. [2003], added specific parameters, such as evaporation, relative humidity (RH) and sea surface temperature (SST), which influence the isotopic composition of the water vapor and, consequently, the rainwater. Evaporation from water bodies includes mainly depleted water molecules, causing the remaining water body to become more enriched, especially over small lakes, where the RH is relatively low and the evaporation rate is larger. As for the SST, the higher the SST, the more intense is the evaporation due to the higher kinetic energy of the individual molecules so that the water vapor is more enriched and so are the resulting raindrops. Another geographical effect that is relevant to the study region is the marine effect, i.e., the time the rain-producing air mass spends over the Mediterranean Sea before reaching the Negev Desert. We suggest that while moving over the sea surface the otherwise continental air (as is reflected by the air trajectories shown in section 3) would gain more heavy isotopes so that it becomes more enriched. This is because the Mediterranean Sea contains mainly heavy stable isotopes [Gat et al., 2003].

[6] The isotopic composition of the rain in Israel has been previously studied by Rindsberger et al. [1983], who analyzed rain samples from 10 rainstorms in its northern part. They found a statistically significant positive linear relationship between the 800 hPa (~2000 m above sea level) temperature at Mid Israel (where soundings station is located) and the isotopic composition, deducing that the latter may be regarded as an indicator for the source latitude of the cold air that produces the winter rains in Israel. Leguy et al. [1983] and Rindsberger et al. [1983] were the first to depict air back trajectory as a basis for identifying the sources and paths of the rain-producing air mass in Israel. The back trajectories derived by Rindsberger et al. [1983] were based on 800 hPa level alone. This approach is limited by its implied neglecting the vertical displacement of air masses, especially in the presence of cyclones, in which the rain is produced. Moreover, the choice of such a midlevel displacement may miss the contribution of the PBL, which is normally confined to the lower levels.

[7] Dody et al. [2010] claimed that to improve our understanding of the fractionation process under atmospheric conditions, sampling in high temporal and spatial resolution is needed. Pfahl and Wernli [2008] discussed the effect of air trajectory on stable isotopes in water vapor in the eastern Mediterranean. They estimated the synoptic scale trajectory, based on the RH, as measured in Rehovot (Israel).

[8] Several studies concentrated on the Negev Desert located at southern Israel. Levin et al. [1980] stated that the isotopic composition is relatively uniform over a wide area on any particular day but differs substantially from one storm to another. Dody [1995] and Adar et al. [1998] showed, for the same geographical area, that the isotopic composition varies sometimes dramatically even within individual rainstorms.

[9] Climatic and synoptic studies have pointed at several sources of air from which the rain over the Negev Desert is associated. Kahana et al. [2002] showed that half of the major floods in the northern part of the Negev are associated with moist air flowing from the Mediterranean and the other half by intrusions of tropical air masses from south. Dayan et al. [2001] and Ziv et al. [2004] found that tropical air masses are involved in autumnal severe floods that occur in south Israel under the influence of a Red Sea trough. Ziv [2001] and Rubin et al. [2007] showed that the moisture in some winter rainstorms transported development produced by tropical plumes [McGuirk et al., 1987, 1988] originating from tropical western Africa.

[10] The linear relationship between the source latitude of the air mass and the isotopic composition of the rain it produces [Rindsberger et al., 1983] suggests that an analysis of the isotopic composition of underground water in North Israel can point at the dominating source of air masses that produced rain in the past. The results of such analysis can therefore serve as a proxy for past climatic conditions in that region. One of our goals is to examine whether this also holds for the central Negev Desert covering the southern part of Israel.

[11] The aim of this study is to isolate the effects expected to influence the isotopic composition of the rain in south Israel based on sampling the rainwater in high temporal resolution, in an order of hours. The factors analyzed are the temperature, amount, marine and continental effects, and an effect related to the height of the convective top proposed here (cloud top effect, see section 3.2). The analysis uses air back trajectories, synoptic maps, and vertical atmospheric cross sections. Section 2 specifies the data sources and methods used. Section 3 elaborates the functional relationships between the isotopic composition and its governing factors and presents three case studies to demonstrate the dynamics involved. Section 4 summarizes results and discusses their implications.

2 Materials

[12] The study region is the central part of the Negev Desert, Israel, having an average annual precipitation of ~100 mm, distributed during 20 rain days during the winter season (on average) [Zangvil and Druian, 1989]. The rain recording and the isotopic sampling were taken from the Sde Boqer station (30.5°N, 34.8°E, see Figure 1). The research covers eight rainstorms that occurred during the years 1990–1993 and described in detail by Dody et al. [2010]. All of the rain samples were analyzed at the International Atomic Energy Agency (IAEA), Vienna. Samples of three rainstorms were also analyzed at the Niedersachsisches Landesmat for Forchung (NLFB) Laboratory in Hannover, Germany. The standard errors for the sampling are ±1‰ for δD and ±0.1‰ for δ18O, which are acceptable for such analysis. No differences were found between the results obtained in the two labs (Dody et al. 2010). In this study, we present only the δ18O values. Details concerning the eight storms analyzed are specified in Table 1.

Figure 1.

Long-term mean annual rainfall map of Israel. The study region is denoted by a six-pointed star.

Table 1. List of the Eight Rainstorms Analyzed and Their Synoptic Features
PeriodNo. of δ18ONo. of Rain SegmentsNo. of Atmospheric SamplesStorm TypeAir Trajectory
25–26 January 199012Not analyzedNot analyzedTropical easterly troughOriginated from Equatorial Africa at the mid levels
1–2 February 1990622Midlatitude low over south Israel, moving eastwardMediterranean
11–13 March 19901533Midlatitude low over Trans-JordanMediterranean
1–2 April 19901544Midlatitude low over Trans-JordanMediterranean at the beginning, later from Europe (through the Mediterranean)
24–25 January 19911823Midlatitude low over Trans-JordanEurope (through the Mediterranean)
7–8 February 1991712Midlatitude low over Trans-JordanMediterranean
5–6 March 19911444Midlatitude low over Trans-JordanMediterranean at the beginning, later from Europe (through the Mediterranean)
23–24 March 19912333Midlatitude low over Trans-JordanMediterranean
Total1101921  

[13] A mechanical continuous sequential rain sampler with high resolution, developed by Adar et al. [1991], was used. This sampler collects 250 to 500 mL of water, equivalent to 1 and 2 mm of rain, respectively. The novelty of this method is that once an assigned volume of water is collected, the weight of the water load in the container is used as a mechanical seal that isolates it as a separate rain sample. The implication is that the higher the rain intensity, the higher the temporal resolution. The sampler has 20 rain bags, which allow the collection of rainwater from a storm with 20 to 40 mm.

[14] Air back trajectories were extracted using the Hybrid Single Particle Lagrangian Integrated Trajectory model of the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (http://ready.arl.noaa.gov/HYSPLIT.php). The displacement of an air parcel is computed based on the 3-D wind field of the National Centers for Environmental Prediction (NCEP)-National Center for Atmospheric Research (NCAR) reanalysis data [Kalnay et al., 1996 and Kistler et al., 2001], given at 2.5° × 2.5° resolution, in 6 h time increments. In the analysis, the levels from which air back trajectories were derived are 500 and 1000 m above ground level (AGL), assuming that these heights represent the PBL from which the rain clouds develop.

[15] The impacts of various effects were examined: the amount effect, the air temperature, the continental and marine effect, and the cloud top effect (described in section 3.2). For each effect, a linear regression with the isotopic depletion was derived, and its statistical significance was assessed through Pearson correlation. The amount effect was analyzed for rain segments each separated by >4 h from each other. The air temperature at the study region was taken from the 850 hPa level, a standard level where data are given. According to Rozanski et al. [1993], this effect holds when the temperature is <15°C. The temperatures observed in this study are <7°C. Taking into account that the typical height of the 850 hPa level is 1500 m, the cloud base is ~1000 m and that the atmospheric lapse rate is <1°C/100 m, the above condition is met in the study area and period.

[16] The continental and the marine effects were estimated by the time the air spent over land and the Mediterranean Sea, respectively, before entering to the study region at 500 and 1000 m AGL (using back trajectories derived from the HYSPLIT software, see above). The top of the convective cloud layer (as an estimate of the cloud top effect, see section 3.2) was estimated by the uppermost pressure level at which the RH exceeded 60%, based on vertical cross sections derived from the NCEP-NCAR reanalyzed data. All above relationships were analyzed applying linear regression.

[17] Two factors were not included in this study. One is the determination of the air origin, which depends on an arbitrary choice of the duration of the back trajectory. Hence, the air origin is treated subjectively and is estimated through the temperature effect (as done by Rindsberger et al. [1983]). The second factor excluded from this study is that of the Mediterranean SST. This is due to its being stable during the months included in this study, January to April, and varies in the small range of 17–19°C.

3 Results

[18] The synoptic features of the eight rainstorms analyzed are specified in Table 1. Seven of them were associated with midlatitude cyclones located over Trans-Jordan, and one, occurred at 25–26 January 1990, was associated with tropical like easterly trough. The rains resulting from the midlatitude cyclones were formed under northwesterly maritime flow that entered the study region from the Mediterranean. Following Kahana et al. [2002] and Saaroni et al. [2010], the clouds in these storms are convective and emanate from the PBL. These rains are enhanced by orography along the northern slopes of the Negev Highland. In contrast to these storms, the rain of 25–26 January 1990 developed within stratified cloud layer centered at 2–3 km height. Because of its exceptional type, source of air, and type of clouds and rain, this storm was excluded from the present study.

3.1 Evaluation of the Factors Affecting Isotopic Composition

[19] The numerical values of the δ18O and the governing factors for the seven storms associated with midlatitude cyclones are specified in Table 2. The linear relationships derived for these factors are presented in Table 3. The relationships for the temperature and the rain amount are shown in Figures 2a and 2b, respectively. Both relations are in agreement with the theoretical expectations (see section 1), but only that of the rain amount is statistically significant (at the 0.9 level, using one directional assumption, although being low, R = 0.38). The continental and the marine effects were found to meet the expected relation but are statistically insignificant (see Table 3). We suggest that the reason for the marginal influence of the marine effect for the study region stems from the fact that in all the samples analyzed, the minimum time spent over the Mediterranean was 15 h, which may be sufficient to gain considerable isotopic enriched composition. It should be noted that when the eighth storm was included in the sample, the temperature isotopic composition relation was inversed (still, statistically insignificant).

Table 2. List of Samples Taken From the Seven Rainstorms Associated With Midlatitude Cyclones and the Values of the Relevant Factors
Serial NumberDateHour (UTC)δ18OT (°C)Time Over Med. (h)Time Over Land (h)Top of Convective Clouds (hPa)
11.2.9021–2.53.5363690
22.2.906–3.52.5486740
311.3.906–1.74486710
412.3.9018–84.54829490
513.3.909–2.54.52646580
61.4.903–4.764626390
71.4.9012–376012680
82.4.903–53157540
92.4.9012–3.54156560
1025.1.9112–51.5333770
1125.1.9118–8.30.5304500
1226.1.9112–9.50203550
137.2.9112–75324400
148.2.9112–53.5704770
155.3.910–58608430
165.3.9112–94288380
176.3.913–51226615
186.3.9112–4.52183780
1923.3.9112–3.87.7205530
2024.3.913–9.56.8189655
2124.3.9118–39209700
Table 3. Linear Relations Between δ18O and its Governing Factors and Their Statistical Significancea
FactorSample SizeLinear RelationshipFit to Theoretical ExpectationsStatistical SignificanceReference
  1. aNote that the cloud top effect is estimated by the pressure level of the cloud tops, so that the positive coefficient means a negative relation of the isotopic composition with the height (in meters).
Temperature21y = 0.26x – 6.26YesNo

Clark and Fritz [1999]

Amount effect19y = –0.15x – 4.3YesAt the 0.9 level"
Marine effect20y = 0.03x – 6.18YesNo

Gat et al. [2003]

Continental effect21y = –0.03x – 5.02YesNo

Clark and Fritz [1999]

Cloud top effect21y = 0.009x – 10.8YesAt the 0.95 level
Figure 2.

δ18O as a function of (a) temperature and (b) rain amount. Each graph contains the linear relationship and R2.

3.2 Hypothesized Additional Effect—The Cloud Top Effect

[20] Here we hypothesize an additional effect related to the height of the convective clouds tops. Since cold air contains depleted composition, the lower top of the clouds implies heavier composition. This idea comes from inspection of individual cases, in which sharp changes in the isotopic composition could not be explained by any of the already known factors. This effect is called hereafter the cloud top effect. The cloud top effect is estimated here by the pressure level separating high values of RH below and low values above (note that higher values of pressure level means lower altitude). The threshold value used here is 60% because in one case, the maximum was <70%. The higher this level is, less portion of cold air entrained with the clouds. The hypothesized relation was examined for our data (21 samples) and found to meet this hypothesis (Figure 3 and Tables 2 and 3) and even found the most statistically significant effect, at the 0.95 level. The correlation between the isotopic composition and the cloud top effect was found to be 0.55. Since the isotopic composition is affected by several factors, this correlation can be regarded rather high.

Figure 3.

δ18O as a function of the level representing the top of the convective layer. The units are hPa. Lower values implies higher level; e.g., 500 hPa is ~5500 m, 700 hPa is ~3000 m, etc.

3.3 Case Studies

[21] This section analyzes three cases that exemplify the impact of the various governing factors on the variations in the isotopic composition. Each case offers a clear example that demonstrates the contribution of specific factors. For each case, the sea level pressure (SLP) maps, the air back trajectories and the rainfall chart of the study region are shown, together with the corresponding δ18O. Note that the three cases are from March and April. This may impose a bias since they represent the end of the rainy season. The only variable that has a seasonal course in the study region is the Mediterranean SST, which remains stable through January–April (see also section 2), from which the rain samples were taken (Table 1).

3.3.1 Case 1: 11–13 March 1990

[22] This case lasted two and half days and yielded 17 mm of rain. The isotopic composition varied between –1.7 and –9.0‰. The dominating synoptic system was a cyclone that was formed in north Saudi Arabia in the morning of 11 March 1990 and remained at that vicinity during the following days. In part of the time, a secondary small-scale cyclone was formed over the Nile Delta, but its effect on the wind over the study area was minor. The resulting wind over the study area undulated around the northwesterly-westerly direction, as is manifested by the air back trajectories noted at three different times during the storm (Figures 4b–4d).

Figure 4.

(a) SLP for 12 March 1990 18 UTC, with the notation L at the centers of the cyclones; air back trajectories arriving at the study region at 500 (blue) and 1000 m (red) AGL for (b) 11 March 1990 04 UTC, (c) 12 March 1990 20 UTC, and (d) 13 March 1990 13 UTC; (e) rain chart Sde Boqer station and (f) the corresponding isotopic composition with notations (red) of the 850 hPa temperature at the study region for selected times.

[23] The storm is divided to three phases according to the rain segments and the origin of the air that entered the study area. At the first phase, the air originated from Turkey (Figure 4b), at the second it originated from Greece (Figure 4c), and at the third from the Eastern Mediterranean itself (Figure 4d). However, the 850 hPa temperature (Figure 4f) was stable during the storm and remained around 4°C, suggesting that the differences among the sources of the air mass affecting the study area were small. In spite of the steady temperature, the δ18O varied considerably, from –1.7‰ at the first phase, to –8‰ at the second and back to –3‰ at the third (Figure 4f). Since the temperature itself did not change during the storm, the changes in the δ18O should be explained by changes in other factors.

[24] At the first phase, the air entered the study region from the sea and then moved ~50 km over land within 6 h (Figure 4b). At the second phase, the continental path of the air was much longer, ~400 km (~24 h span, Figure 4c), presumably due to a secondary cyclone that was formed over the Nile delta (Figure 4a). This can explain why the isotopic composition was more depleted than in phase 1. At the third phase, the air path could not be determined because of the high differences between the back trajectories entering the study region at 500 and 1000 AGL (Figure 4d).

[25] Two additional factors can better explain the δ18O sharp changes of the isotopic composition during the storm: the amount effect and cloud top effect. The rainfall in the first phase was 4 mm; at the second, 12 mm; and at the third, 2 mm, in agreement with the changes in the isotopic composition, –1.7, –8.0, and –2.5‰, respectively. As for the top of the convective layer, it rose from 710 hPa (~2900 m) in the first phase to 490 hPa (~5700 m) in the second and subsided back to 580 hPa (~4400 m) at the third (for more details, see Table 2). Note the positive relation between the pressure level of the convective layer top and the δ18O (Figure 3). This case demonstrates the contribution of the continental, amount, and the cloud top effect.

3.3.2 Case 2: 1–2 April 1990

[26] This case lasted two days and yielded 20 mm of rain. Figure 5a shows the synoptic situation as represented by the SLP of 2 April 1990 00 UTC, showing a pronounced cyclone centered over northwest Iraq. The resulting flow over the study area was northwesterly during the entire storm. The back trajectories, taken at two different times (1 April 1990 14 UTC and 2 April 04 UTC, Figures 5b and 5c, respectively) show that the air that entered the study region at the lower levels, represented by the heights of 500 and 1000 m AGL, approached from the northwest. Figure 5d indicates that the rain was concentrated in four distinct segments, and 70% of it was obtained in the second and the third of them, 8 and 6 mm, respectively.

Figure 5.

(a) SLP for 2 April 1990 00 UTC; air back trajectories arriving at the study region at 500 (blue) and 1000 m (red) AGL for (b) 1 April 1990 14 UTC and (c) 2 April 1990 04 UTC; (d) rain chart Sde Boqer station and (e) the corresponding isotopic composition with notations (red) of the 850 hPa temperature at the study region for selected times.

[27] Comparison between the first phase of the storm, including the first two rain segments, and the second, including the last two segments (see Figures 5d and 5e), indicates that two factors contributed to a depletion of the isotopic composition, while two others contributed to its enrichment. One depleting contribution was a change in the source of the air producing rain (compare Figures 5b and 5c). The back trajectories corresponding to the first phase (Figure 5b) originated from the Mediterranean region, south of Greece, and that in the second phase originated farther north, from the Black Sea (Figure 5c). The temperature dropped, accordingly, from 6–7°C in the first phase to 3–4°C in the second. Another depleting contribution originated from the continental effect; the span of the air over land in the first phase was 12–26 h and dropped to 7 h in the second (Table 2). At the same time, other factors acted at the opposite sense, i.e., to enrich the isotopic composition. One is the cloud top effect, expressed by a drop of the top of the convective layer from 390 hPa (~7500 m) to around 600 hPa (~4000 m, Table 2). The second is the marine effect; the time the air spun over the Mediterranean was ~50 h at the first phase but only 15 h in the second.

[28] This case demonstrates how two effects, the temperature and the continental, acting at one direction, are counteracted by the cloud top and the marine effects. It should be noted that beside the small changes found among the average values of the isotopic composition for the different rain segments the differences among the individual samples were larger (in the range of –2.2‰ to –7‰) but were smoothed out due to the limited resolution of the synoptic data.

3.3.3 Case 3: 5–6 March 1991

[29] This case lasted two days and a half and yielded 17 mm of rain. The synoptic dominant feature was a midlatitude cyclone that was formed in north Iraq on the morning of 5 March 1991 and remained there for the rest of the storm (exemplified by Figure 6a), with only slight changes. A northwesterly flow dominated the lower levels in the study region, as is reflected by the air back trajectories (Figures 6b and 6c). The origin of the air that entered the study area at the beginning of the storm was the Mediterranean Sea (Figure 6b), then shifted to the Black Sea (Figure 6c) and later to the Balkans (not shown).

Figure 6.

(a) SLP for 6 March 1991 00 UTC; air back trajectories arriving at the study region at 500 (blue) and 1000 m (red) AGL for (b) 5 March 1991 01 UTC and (c) 5 March 1991 12 UTC; (d) rain chart Sde Boqer station and (e) the corresponding isotopic composition with notations (red) of the 850 hPa temperature at the study region for selected times.

[30] An examination of the amount effect (compare Figures 6d and 6f) shows that during the most intensive rain segment (the second, with 7 mm of rain), the isotopic composition was the depleted, in agreement with the expected relation. On the first day of this storm, the temperature isotopic composition relationship is clearly seen. While the 850 hPa temperature dropped from 8 to 4°C, the δ18O became depleted from –4 to –8‰. From the beginning of 6 March, while both the back trajectories and the temperature remained unchanged (for the temperature, see Figure 6e), the δ18O became considerably enriched, from –9 to –4‰. The only factor that changed considerably during 6 March was the cloud top effect. This is represented in Figure 7, showing a vertical cross section of the RH along 35°E. It shows that at 5 March 18 UTC the convective layer, within which the RH exceeded 60% at 30.5°N (where the study area is located), extended up to ~400 hPa, implying that the cloudiness could have a depth of ~7 km. Figures 7b–7d show that the moist layer became thinner gradually; until at 6 March 12 UTC, the end of the rainstorm (Figure 6d), its depth was <3 km (Figure 7d). The decrease in the depth of the convective layer during 6 March explains the gradual enrichment of the isotopic composition during the second day of the storm, in spite of the steadiness of the other atmospheric factors. This case clearly demonstrates the contribution of the cloud top effect.

Figure 7.

Vertical cross section through the 35°E longitude of the relative humidity for (a) 5 March 18 UTC, (b) 6 March 00 UTC, (c) 6 March 06 UTC, and (d) 6 March 12 UTC. The vertical line denotes the latitude of Sde Boqer.

4 Discussion and Summary

[31] The study examines the δ18O of rain observed in the center of the Negev Desert, Israel. The results of seven rainstorms, all characterized by convective clouds associated with midlatitude cyclones found over Trans-Jordan, were analyzed using synoptic maps, vertical atmospheric cross sections, and air back trajectories, to identify the major factors determining the isotopic composition of the rainwater. The rainwater sampling was done in high temporal resolution, i.e., a time scale of hours. The reconstruction of air back trajectories is based on the HYSPLIT model.

[32] Five factors were addressed here (Table 3). The first is the temperature effect, which is a proxy for the origin of the air mass within which the rain clouds are formed; the second is the amount effect; the third is the marine effect, represented by the time the air spent over the Mediterranean Sea before reaching its southeastern coast; the fourth is the continental effect imparted by the passage of the rain-producing air over land before reaching the study area. The fifth effect is proposed here, entitled the cloud top effect, related to the vertical depth of the convective cloud layer.

[33] The temperature effect, expressed by a positive relation between the lower level temperature and the δ18O (found for north Israel by Rindsberger et al. [1983]), was also found in this study, although not statistically significant, presumably due to the interference with other factors. The amount effect, expressed in a negative relation between the rainfall at a certain rain segment and the isotopic composition, was found here to have low correlation (0.38) and statistically significant, at the 0.90 level. The marine effect, expressed in a positive relation with the isotopic composition, and the continental effect, expressed in a negative relation with the isotopic composition, were both identified in this study, although being statistically insignificant.

[34] The cloud top effect, proposed here, assumes that the lowering of the top of the convective layer prevents the entrainment of upper level cold air, containing depleted isotopic composition, into the clouds. In other words, the truncation of the cloud tops implies a smaller portion of light isotopes, i.e., enriched δ18O. The cloud top effect is exemplified here mostly by case 3, above, in which a substantial enrichment of the δ18O was observed while minimal change in the other effects was observed, except for a considerable lowering of the top of the convective layer. The statistical significance of this effect was found to be the highest (at the 0.95 level), and the correlation R was 0.55.

[35] The cloud top effect and the amount effect are expected to be correlated since as the cloud tops became higher, they tend to produce more rain. Unfortunately, these two effects could not be correlated due to the different rain sampling methodologies (see section 2).

[36] It should be noted that this study covers only part of the rain-producing systems, i.e., the midlatitude cyclones while positioned over Trans-Jordan. These produce rain showers associated with convective clouds that develop within the PBL. For the sake of homogeneity, we selected out one additional storm (No. 8) associated with medium stratified clouds that developed within tropical air mass. Storms associated with tropical air masses, which contribute considerable rain in the Negev Desert (e.g., Kahana et al. [2002]) should be studied separately, after a considerable amount of rain samples is accumulated.

[37] Our results (Table 3) indicate that for the Negev Desert of Israel, the monotonic relationship between the isotopic composition and source latitude of the rain-producing air mass, as found by Rindsberger et al. [1983], exists. However, this relation is not statistically significant, presumably due to the interference with the other effects, such as the cloud top effect, which was found highly significant. It is possible that in the Negev Desert, the changes in the top of the convective layer are associated with the proximity of this region to the descending branch of the Hadley Cell, which may lower it more frequently than in North Israel, where the temperature effect was found statistically significant. Moreover, the northern part of Israel obtains its rain mainly under southwesterly winds [Saaroni et al., 2010] entering from the Mediterranean Sea. This implies that the probability of the rain-producing air to cross a considerable continental area (and hence to be subjected to the continental effect) is low (Goldreich, 2003).

[38] An attempt to evaluate the variability of the isotopic contribution explained by the above effects was done by applying a multi regression analysis. The analysis does not include the amount effect due to the inability to unify the latter data set, based on the spatial distribution of the rain segments, with the other samples, based on the availability of atmospheric data (in 6 h intervals). The multiregression analysis (using the “stepwise” option) identified only the cloud top effect as significant.

[39] The effect found most significant in this study, the cloud top effect, together with the temperature effect (statistically significant at the 0.85 level), reflect similar atmospheric situation, i.e., a southward intrusion of polar air and an equatorial retreat of the descending branch of the Hadley Cell. If the rain over the study area is found to be contributed mainly by midlatitude cyclones, then the isotopic composition of δ18O in the underground water in that region would serve as a proxy for past climate as it reflects the degree at which the midlatitude systems affect the Negev Desert.

Acknowledgments

[40] This study was supported by the Israeli Science Foundation (ISF grant 108/10). Thanks also are given to Mrs. Cathy Kelly for reviewing the paper.

Ancillary