Stable Isotope Composition of Cyclone Mekunu Rainfall, Southern Oman

Cyclone Mekunu hit the southern Arabian Peninsula in late May 2018 and brought rainfall amounts that accounted for up to 6 times the mean annual precipitation. Coming from the Arabian Sea, a quite underdocumented region with regard to cyclones, the storm eye crossed the Omani coast approximately 80 km east of the border to Yemen. Using automatic samplers, rainfall samples were collected during the event at three locations along a transect almost parallel to the storm track. The stable isotope analyses show a wide range of δ values, with minimum and maximum values of −17.01‰ δ18O and −1.77‰ δ18O and −122.2‰ δ2H and −1.6‰ δ2H. On average, rainfall becomes isotopically lighter with elevation, but rather irregularly. In view of high wind speeds probably precluding a gradual rainout of ascending air masses, a “pseudo elevation effect” seems likely. Our measurements expand the known δ value range of local cyclones by about 6‰ for δ18O and by nearly 50‰ for δ2H. The isotopic composition of the annual Indian Summer Monsoon shows values of −0.93‰ δ18O to 2.21‰ δ18O and −2.1‰ δ2H to 23.7‰ δ2H. Thus, there is a clear difference in the dual isotope signatures of the two precipitation systems in the area. Our findings enable an assessment of the impact of cyclones on the hydro(geo)logical system. For the arid Najd area, we demonstrate that the isotopic signatures of groundwater samples fall between those of cyclone and (paleo)monsoon precipitation, suggesting that several rainfall types may have contributed to replenishment.

While the method is generally elegant and simple, it requires a sound knowledge of the isotopic composition of precipitation. Seasonal effects can often be evaluated with monthly long-term data from the Global Network of Isotopes in Precipitation (GNIP; IAEA/WMO, 2020). Yet, studying the relevance of extreme events, such as tropical storms (depending on the region called cyclones, hurricanes, or typhoons) for recharge is usually challenging. The main reason is that the corresponding isotope data are limited. This, in turn, is related to the often irregular and sudden occurrence of these storms, their large spatial extents, and the difficulties in predicting their course. Targeted sampling is therefore challenging. Increased attention in the scientific community, combined with unconventional approaches (e.g., public involvement in sampling) and development of sampling technology, has enabled new data for different regions, including Australia (Munksgaard et al., 2015), China (Xu et al., 2019), Costa Rica , Japan (Fudeyasu et al., 2008;Ohsawa & Yusa, 2000), Puerto Rico , and the United States (Good et al., 2014;Lawrence et al., 1998). These studies show that storm precipitation often exhibits (1) a significant depletion in heavy isotopes and (2) large intra-event δ value variations. Yet clearly, there is still a data gap concerning tropical cyclones . The number of studies on this topic is still small, and there are regions with little or no information available. The Arabian Sea, presented here, is one of them.
The greater Salalah area in Southern Oman ( Figure 1) experiences an annual monsoon (locally called khareef) and occasional cyclones that can deliver rainfall amounts that substantially exceed mean annual precipitation. To the best of our knowledge, the study area is the only place on the Arabian Peninsula influenced by both precipitation systems. Since the monsoon occurs every summer in the form of a quasi-continuous drizzle (unlike many other monsoons) and plays an important role for water supply, ecosystems and forestry, livestock grazing, and tourism (Abdul-Wahab, 2003;Friesen et al., 2018;Hildebrandt et al., 2007;Shammas, 2007), it has received considerable attention by the scientific community. Apart from seasonal precipitation amounts (typically 55 and 220 mm in the coastal plain and the Dhofar Mountains, respectively; Hildebrandt et al., 2007), also the isotopic composition of the monsoon rain has been studied. With δ 18 O values from −0.93‰ to 2.21‰ and δ 2 H values between 2.1‰ and 23.7‰ (daily samples; n = 153; Clark, 1987;Strauch et al., 2014;Wushiki, 1991), it was found to be similar to the isotopic composition of seawater. However, corresponding data on cyclones in the region are scarce and we lack an overview of possible δ value ranges and spatial heterogeneities. Whether cyclones show a classical elevation effect (depletion in heavy isotopes at greater elevation) is not known either. In part, this data scarcity is caused by the lack of a long-term isotope monitoring network in Oman (Weyhenmeyer et al., 2002) and also the irregular occurrence of the cyclones (every 2-6 years; Friesen et al., 2018) likely plays a role. For southern Oman five cyclone rainfall values have been reported so far, four for Cyclone O6-A in 1992 and one for Cyclone Keila in 2011 (Macumber et al., 1995;Strauch et al., 2014 taken on the fifth day of the event and is even more depleted in heavy isotopes (−10.76‰ δ 18 O and −73.0‰ δ 2 H). All samples were taken as grab samples (i.e., not as bulk samples integrating over the entire event or a specified time period) and corresponding rain amounts are unknown. Hence, these isolated samples do not allow to trace the isotopic development, to calculate Event Meteoric Water Lines, or to provide a precipitation-weighted mean isotopic composition. Data on cyclone-triggered surface runoff, sampled by the same authors, is similarly difficult to interpret.
In May 2018, Cyclone Mekunu hit the region and brought more than 500 mm of rain within a few days (Government of India, 2018; PACA, 2018). It caused devastating floods, 31 casualties, and damage exceeding 1.5 billion US$ (AON, 2018). However, it also offered the possibility to sequentially collect rain samples for isotope analyses along a transect from the coast to the Dhofar Mountains. Such data may help to clarify the general significance of extreme events on groundwater recharge. This is especially important for the area north of the Dhofar Mountains, the arid Najd, which is Oman's main agricultural area and experiences slow depletion of its water resources (Saber et al., 2017). While there is consensus that the groundwater, occurring at depths of up to 700 m, has been recharged under more humid paleoclimatic conditions (in the Early to Middle Holocene and Late Pleistocene), the mechanisms are still under debate. Some researchers favor the hypothesis that replenishment was mainly caused by a stronger (Indian Ocean) monsoon (Fleitmann et al., 2003;Matter et al., 2015;Rosenberg et al., 2011) that reached the interior of the Peninsula due to a northward shift of the Intertropical Convergence Zone (Burns et al., 1998. Others believe that (cyclonic) storms were the dominant source (Al-Mashaikhi et al., 2012;Clark, 1987;Clark et al., 1987). Stable isotopes have the potential to resolve this issue (Meredith et al., 2018;Sultan et al., 2008), but we need data for both rainfall types, i.e., monsoon and cyclone precipitation.
Since the lack of reliable data on the latter represents an important knowledge gap, our main questions are as follows: To what extent do local cyclone rains differ from monsoon precipitation isotopically? Do local cyclone rains show the classic elevation effect, that is, a heavy isotope depletion at greater elevation? Have cyclones been a major contributor to the Najd groundwater system in the past? Since the hydrochemical composition of cyclonic rain in the region is also largely unknown, we complement our unique isotope data with major ion analyses, which will be useful for future groundwater quality studies.

Cyclone Mekunu
Cyclone Mekunu formed over the Arabian Sea on 20 May 2018. Starting as a low-pressure area, the system evolved within the next 5 days into an event of the category "Extremely Severe Cyclonic Storm" (defined by wind speeds between 166 and 220 km h −1 ; Government of India, 2018) with a minimum central pressure of 960 mbar. During its formation, promoted by high sea surface temperatures of 29°C to 31°C (Government of India, 2018), the system moved westward and then turned in a north-northwestern direction (translational speed of 10.4 km h −1 ). The center of the storm crossed the Omani coast approximately 45 km west of the city of Salalah around midnight on 25/26 May. At this time, the system had a diameter of several hundred kilometers ( Figure S1). Continuing north-northwestward, the system moved inland and reached the border of Oman on 27 May where it finally dissipated ( Figure 1a). The track length of the cyclone was approximately 1,400 km (Government of India, 2018). Air mass back trajectory modeling (HYSPLIT; Stein et al., 2015;Rolph et al., 2017) suggests that contributing air masses largely originated from the Western Arabian Sea ( Figure S2).
Weather station data indicate that rainfall occurred along the entire coast of Dhofar, which corresponds to a corridor of about 400 km. Rainfall was observed in the coastal plain and the Dhofar Mountains and also reached the Najd desert north of the mountains. In Salalah, 617 mm of rain fell between 23 and 27 May (Government of India, 2018). This is 5 to 6 times the mean annual precipitation of Salalah.

Methods
Three automatic rain collectors were installed in the coastal plain (Station M1) and the mountains (Stations M2 and M3), to measure precipitation amounts and gather samples (Table S1). To avoid postsampling evaporation, the devices harness the "tube-dip-in-water collector with pressure equilibration tube" principle (Gröning et al., 2012) and automatically collect samples in predefined time intervals (for details, see Michelsen et al., 2019). The rain was collected via a funnel and drained freely in the sampling device below. The funnel diameter was 7.5 cm at Stations M1 and M3, and 14.5 cm at Station M2. At station M1, the collector was mounted on a roof (approximately 10 m above ground). At Stations M2 and M3, the funnel was installed at a height of 2 m. All collectors started sampling on 24 May at 04:00 p.m. Integral (i.e., cumulative) samples at 4-h intervals were taken until 12:00 p.m. on 27 May. For logistical reasons Interval 18 lasted 28 h until 28 May 04:00 p.m. From then until 29 May afternoon, when the collectors were removed, the samples were again collected at 4-h intervals. In total, this provided 23 intervals for sampling.
Station M1 was located directly at the shoreline and thus almost at sea level (10 m amsl). The stations in the mountains were at 510 m amsl (M2) and at 880 m amsl (M3), almost at the top of the Dhofar Mountains ( Figure 1b). The transect length was 30 km in total, 21 km between M1 and M2 and 9 km between M2 and M3.
After dismantling the collectors, the HDPE-bottles were sealed with their original caps and electrical tape. None of the bottles had overflowed. On the next day, the samples were transported to Germany and measurements started 5 days later. The stable isotope composition of water (δ 18 O, δ 2 H) was measured by laser spectroscopy using isotopic liquid water analyzers (L2120-i by Picarro and TIWA-45EP by Los Gatos Instruments). We calibrated the analytical devices with the two extremes High ( (Coplen & Wassenaar, 2015). The results were expressed in per mil (‰) using the conventional delta-notation relative to V-SMOW. The 1σ reproducibility is commonly better than ±0.2‰ (δ 18 O) and ±0.4‰ (δ 2 H) for both instruments.
A handheld unit was used for electrical conductivity (EC) measurements (Multi 3430 by WTW). Major ion concentrations were analyzed by ion chromatography (882 Compact IC plus by Metrohm) with a precision better than ±3%.

Rainfall Distribution
During Cyclone Mekunu, rain was collected for six consecutive days at all three sampling locations ( Figure 1b and Table 1).
When the first rain showers occurred on land on 24 May, the storm eye was about 330 km to the south of the Salalah coastal plain. The largest rain amount for the whole 6-day period, a total of 745 mm, was measured at station M3. At stations M1 and M2, the total amounts were much smaller, 298 and 242 mm, respectively. The highest daily amounts occurred on 25 May. The values recorded at M3 (398 mm) and M1 (171 mm) exceed the corresponding mean annual precipitation amounts (based on data from 1992-2014, see Friesen et al., 2018). From 25 May 04:00 p.m. to 26 May 04:00 a.m. (when the storm made landfall; Intervals 7, 8, and 9), 384 mm of precipitation were measured at station M3 and 117 mm at station M2. In both cases, this is about 50% of the total event precipitation. At station M1, about 125 mm were recorded in this phase, corresponding to approximately 40% of the total event rainfall. During this period, the eye of the storm was closest to the sampling stations (approximately 70 km, Figure 2).
Rainfall amounts first increased at the coastal station M1 (slightly before M2 and M3), but the maximum coincided with landfall (Interval 7; 25 May 04:00 to 08:00 p.m.). A second, much smaller peak was recorded at all stations during the night of 26/27 May (Figure 2).
At station M3, rain occurred during each interval (n = 23); at stations M2 (n = 20) and M1 (n = 13) there are periods with no rain (Table 1 and Figure 2). Although rainfall was observed for each interval at the highest station M3 and almost all intervals at station M2, there is no overall trend of increasing rain amounts with elevation. The corresponding order for the total rainfall is M3 at 880 m amsl (745 mm), M1 at 10 m amsl (298 mm), and M2 at 510 m amsl (242 mm). This is interpreted as a typical characteristic of storm events with a heterogeneous rainfall distribution (e.g., due to spiral rain bands and the less rainy areas between them; see Figure S1). These features and the general movement of the system lead to significant spatiotemporal heterogeneities of rainfall amounts, a characteristic that also has been observed in other cyclone studies (Good et al., 2014).

Water Resources Research
The last coordinates for the storm center are given for the morning of 27 May (Interval 15) with the center around 280 km to the northwest of the sampling stations (Government of India, 2018). Even though the low-pressure area was not classified as a cyclone anymore, rain occurred for another 3 days (Intervals 16 to 23).

Rainfall Stable Isotope Composition
Stable isotope signatures are available for 8 (M1), 17 (M2), and 20 (M3) of the 23 intervals (Table 1). The smallest number (8) also represents the number of intervals for which data are available for all three stations: Intervals 4,5,6,7,8,14,15,and 18 (Figure 2 (Table 1). The observed pattern indicates a relation between the distance to the cyclone eye, the rain amount, and its isotopic composition ( Figure S3). With decreasing distance to the storm eye and thus to the associated eye walls that contain the moisture (see also Figure S1), rain amounts increase and, in turn, the isotopic composition becomes more depleted in heavy isotopes. The reverse happens with increasing distance when the storm moves away from the area.
The generally light isotopic signatures of our samples are attributable to efficient vapor condensation, which is generally promoted by the size and longevity of cyclonic events, as well as their great cloud heights ; here about 15 km; Cardellach et al., 2019). Moreover, isotopic exchange between

Water Resources Research
falling rain and converging vapor frequently plays a role Munksgaard et al., 2015). The encountered periodic pattern, in turn, is associated with spiral rain bands passing our sampling locations (e.g., Munksgaard et al., 2015; see also Figure S1). The theoretical background of this trend was calculated by Lawrence and Gedzelman (1996) applying a Rayleigh-type distillation process with progressing rain bands.
In dual isotope space, our data points follow the Global Meteoric Water Line (GMWL; Craig, 1961; see Figures 3 and S4-S6), resulting in a moderate deuterium excess (d) range (d = δ 2 H-8 δ 18 O; Dansgaard, 1964), between 8‰ and 15‰. Moreover, the data partly match the aforementioned wide δ value ranges for cyclonic storms in the region (Macumber et al., 1995;Strauch et al., 2014), but also expand the known range by about 6‰ for δ 18 O and by nearly 50‰ for δ 2 H (Figure 3). The plot also reveals that Cyclone Mekunu showed large intra-event δ value variations, confirming observations for cyclonic storms elsewhere (Gedzelman & Lawrence, 1982;Munksgaard et al., 2015).   (2) that the prevailing wind speeds (up to 175 km h −1 ) do probably not allow for a regular gradual rainout of air masses ascending the mountains. Hence, the effect is more likely to represent a "pseudo elevation effect." Overall, our storm samples confirm the isotopic difference between cyclones and the annual monsoon postulated in previous studies based on a much smaller and partly erratic data set (few grab samples; Macumber et al., 1995;Strauch et al., 2014). Although a few of our samples are only slightly depleted in heavy isotopes, there is no overlap of monsoon and cyclone precipitation (Figure 3)-although both weather systems have their origin in the Arabian Sea (Figures 1a and S2).
The reason is twofold. First, the size, longevity, and cloud heights of Cyclone Mekunu cause an efficient vapor condensation, leading to an isotopically light composition (see above). Second, the annual monsoon exhibits a unique isotopic signature as well. It shows a small δ value scatter, with δ 18 O and δ 2 H values ranging from −0.93‰ to 2.21‰ and from −2.1‰ to 23.7‰, respectively (n = 153; Clark, 1987;Strauch et al., 2014;Wushiki, 1991). The reason for this small scatter and the similarity to the seawater signature (somewhat untypical for monsoon rainfall) is a relatively short atmospheric residence time and a first-stage condensation process (Clark, 1987).

Chemistry
The hydrochemical data obtained for the Cyclone Mekunu samples are presented in Tables S2-S4. The EC was highly variable, ranging between 6 and 375 μS cm − 1 . The EC was highest for the coastal station M1 and was often associated with small precipitation amounts (Table S2 and Figure S7). The remarkable salinity range and the tendency of higher solute concentrations at station M1 (especially Na + and Cl − ) are also reflected in Figure 4a. Since the data points plot on or close to the seawater dilution line, it seems likely that the two ions have a sea spray origin (note wind speeds of up to 175 km h − 1 ; Government of India, 2018). Yet, not all ions are derived from sea spray. When plotting Ca 2 + against Cl − , for instance, the bulk of the data points lie above the seawater dilution line (Figure 4b). Hence, the samples have a higher concentration of Ca 2 + than one would expect based on the corresponding Cl − value and an assumed ion ratio resembling that of seawater (Ca 2 + excess). This excess is particularly high for the inland stations M2 and M3 and suggests an additional Ca 2 + source (apart from seawater). As SO 4 2 − concentrations were also higher ( source. In the study area, these minerals occur in the Eocene Rus formation cropping out north of the Dhofar Mountains (Beydoun, 1966). However, also salt pans that cover large parts of the southeastern Arabian Peninsula (approximately 36,500 km 2 ; Schulz et al., 2015; see Figure 1a) contain these minerals. As they are known to be eroded by wind and have been traced in precipitation of the region , they represent another potential source.
Nearly all points in the Ca 2 + versus SO 4 2− plot lie above the line representing gypsum/anhydrite dust dissolution ( Figure 4d). This considerable Ca 2 + excess (over SO 4 2− ) demonstrates that there is an additional source, most likely calcite dust (CaCO 3 ). A probable candidate in this context is the limestone-dominated Umm Er Radhuma formation (Paleocene to Eocene) with large outcrops in the area, for example, at the southern flanks of the Dhofar Mountains (Beydoun, 1966;Müller et al., 2016; see Figure 1b).
While the postulated lithogenic imprint by gypsum/anhydrite and calcite is visible in samples from all stations, the two inland stations M2 and M3 are, not surprisingly, more affected (higher excess of Ca 2 + and SO 4 2− ) than the coastal station M1. The latter, in turn, is more influenced by sea spray (see also Figure S9).
Interestingly, elevated EC values and ion concentrations do occur at the beginning of the cyclone event but are not limited to the first intervals (cf. Michelsen et al., 2015). Instead, they also appear later, for instance, on 26 May, directly after the first rainfall peak associated with the landfall (Figures S7 and S8). For this phase, one can assume a rather wet terrain, possibly hampering local wind erosion. Hence, the elevated concentrations at late stages of the event might indicate an influence by soluble material mobilized further away and thus a larger geochemical footprint of the cyclone.

Implications for (Paleo)groundwater Recharge
In terms of the Najd groundwaters, both precipitation systems (monsoon and cyclone) are, in principle, recharge candidates ( Figure 3). Yet, due to the great residence times of the groundwater (often thousands or ten thousands of years; Clark, 1987;Müller et al., 2016), the ancient equivalent of the monsoon (called paleomonsoon hereafter) has to be taken into account as well. In the literature, the Early to Middle Holocene (10-6 ka BP) is frequently reported as a more humid period in Southern Arabia Fleitmann et al., 2003;Fuchs & Buerkert, 2008), but also parts of the Late Pleistocene (e.g., 80-75 ka BP; Fleitmann et al., 2011;Rosenberg et al., 2011).

Water Resources Research
Hence, we have to consider three isotopically distinct recharge sources for the Najd groundwater.
The Najd groundwater δ 18 O and δ 2 H values range from −6.2‰ to −0.1‰ and from −48‰ to 5‰, respectively, and show no influence of evaporation during infiltration (Al-Mashaikhi, 2011). They plot within the isotopic signatures of all three potential recharge sources ( Figure 5).
Hence, the groundwater could be a (paleo)monsoon-cyclone mix that can be described with the following mixing model.
where δ Mix is the apparent mixed isotope composition, δ (Paleo)monsoon is the isotopic composition of the paleomonsoon or the monsoon, and δ Cyclone is the composition of the cyclone (δ 18 O). f (Paleo)monsoon and f Cyclone are the corresponding recharge fractions.
If 35% cyclone recharge are mixed with 65% monsoon recharge, the resulting water matches the arithmetic mean of the Najd groundwater samples of −4.57‰ δ 18 O and −29.73‰ δ 2 H. The isotopically lightest groundwater signature of −6.48‰ δ 18 O and −47.2‰ δ 2 H, can be mimicked by choosing proportions of 52% and 48%, respectively. Although such apparent mixing ratios might seem plausible in view of Figure 3 and a recent cyclone indeed led to recharge in the Najd -Müller (2012) reports a groundwater level increase by >4 m during Cyclone Keila (November 2011), they are not appropriate for past recharge scenarios. Taking into account the long groundwater residence times, mixtures of cyclone and paleomonsoon waters have to be considered. In this case, significantly lower proportions of the cyclone water (7% and 30%) are required to match the isotopic composition of the groundwater (to obtain the mean and the lightest groundwater signature, respectively; Figure 5). Hence, this scenario does not require a major cyclone contribution to the recharge of the Najd groundwater.
Apart from groundwater, our findings might also be relevant for studies of high-resolution paleoclimate archives. The isotopic signature of cyclones may be recorded in freshwater mollusks (Lawrence, 1998) or tree rings (Slotta et al., 2019). Speleothems are particularly noteworthy archives on the Arabian Peninsula since much of the paleoclimatic framework is based on data from Omani caves (Burns et al., 1998Fleitmann et al., 2003Fleitmann et al., , 2004

10.1029/2020WR027644
Water Resources Research δ 18 O signature of speleothems via the amount-effect (Dansgaard, 1964), this approach might partly be oversimplified, because microphysical effects and cloud type apparently exert comparable influences on the rainfall's isotopic composition (Konecky et al., 2019). Hence, isotopic data on modern tropical cyclones may represent crucial benchmarks with the potential to improve our understanding of the underlying relations and interplays. Given the general possibility of rapid cyclone imprints on speleothems (Frappier et al., 2007) and the presence of actively growing speleothems in the study area (Fleitmann et al., 2004), it seems timely to study local speleothems to trace Cyclone Mekunu (and similar events) in this archive. This could help to understand the current impact of rain events on the underground (groundwater and speleothems) and, building on it, to draw conclusions about the occurrence and impact of storm events in the distant past.

Conclusions
Given that previous data on cyclone precipitation in southern Oman only comprised six values, our 45 samples from the year 2018 represent a much more solid database. They expand the known δ value range by about 6‰ for δ 18 O and by nearly 50‰ for δ 2 H.
The strong spatiotemporal variability during the event underscores the need for sequential sampling and the calculation of precipitation-weighted means (instead of grab sampling and arithmetic means; cf. Macumber et al., 1995;Strauch et al., 2014). In principle, more stations would be helpful as well, but their number is often limited by resource constraints and safety issues.
The precipitation-weighted means decrease with elevation, but rather irregularly. Given the high wind speeds probably preventing a gradual rainout of ascending air masses, a "pseudo elevation effect" seems likely. The large number of samples in combination with rainfall amounts also allows us to calculate an overall weighted mean composition for Cyclone Mekunu (around −11‰ for δ 18 O and −80‰ for δ 2 H).
All values from 2018 confirm more negative δ values for storm precipitation compared to monsoon rain, which is today's main precipitation source in southern Oman. Although a few samples are only slightly negative, there is no overlap between monsoon and cyclone precipitation. This observation thus also supports the results of previous cyclone studies in other areas, where little or no overlap between cyclonic storms and the background signal (usual precipitation) was observed.
Results of the hydrochemical analyses revealed that all samples were freshwater (max. EC 375 μS cm −1 ) but showed variable compositions (i.e., no unique fingerprint), with a dominant sea spray impact at the coast and stronger lithogenic imprints (gypsum/anhydrite and calcite) further inland.
Our findings may be useful for a number of related fields, including hydro(geo)logy. We could for instance show that the isotopic signatures of groundwaters from the nearby Najd area fall between those of the cyclone and (paleo)monsoon rains, which suggests that several precipitation types could have contributed to recharge.

Data Availability Statement
The precipitation stable isotope and hydrochemistry data are provided in the supporting information and are archived in the PANGAEA database (https://doi.pangaea.de/10.1594/PANGAEA.909045).