Rain samples were collected for isotopic analyses during the entirety of an extreme rainfall event in Beijing, China, on 21 July 2012, the city’s heaviest rainfall event in the past six decades. Four stages of the storm event have been identified with corresponding isotopic characteristics: 1) isotopes deplete as rain increases, 2) isotopes enrich as rain decreases, 3) isotopes quickly deplete as rain increases, and 4) isotopes remain constant as rain reduces to a small amount. The rainout effect dominates the depletion of isotopic composition in stages 1 and 3. The incursion of a new air mass with enriched heavy isotopes was the main cause for the enriched isotopic composition during stage 2. A Rayleigh distillation model was used to describe the isotopic trends during stages 1 and 3. The Rayleigh distillation model and a binary mixing model were used to estimate the initial isotopic composition of different air masses, which were found to be similar to δ18O of precipitation at nearby Global Network of Isotopes in Precipitation stations representing southwest and southeast trajectories. The results are in agreement with meteorological arrays analysis. This model also indicates that 29% of the initial vapor from the southwest trajectory was precipitated in stage 1, followed by a mixing process between southeast and southwest moisture. In stage 3, up to 56% of mixed moisture was precipitated, among which ~65%–100% was from southeast moisture.
There is presently a lack of research in the area of connecting atmospheric precipitation to its specific source water, both generally and in relation to extreme rain events. While plenty of general circulation models (GCMs) have been explored for tracing water vapor (e.g., Koster et al. 1986; Numaguti 1999; Bosilovich et al. 2003), that approach is overly sensitive to the model’s time–space resolution and the validity of the parameterization schemes used. Furthermore, it is not possible through this approach to estimate the relative contribution of different air masses to the precipitation. Previous studies have used trajectories to examine and quantify the pathways of the air masses that produced the precipitation (e.g., D’Abreton and Tyson 1996; Crimp and Mason 1999; Stohl and James 2004). But these trajectories have been calculated based on wind patterns derived from meteorological analysis and parameterizations for turbulence, and the criteria selected to gather the calculated trajectories differ for different people. Therefore, it is necessary to develop other independent methodologies to verify the model-based results.
Isotopes in precipitation can be effective for identifying and quantifying source areas of precipitation. It is well established that isotopes in precipitation can be used as a tracer to detect water vapor sources and airmass transport pathways, which form the basis for our understanding of the mechanisms of atmospheric moisture formation and transport over the time frame from months to years (Breitenbach et al. 2010; Kong et al. 2013; Li et al. 2014; Tan 2014; Welker 2000). Araguas-Araguas and Froehlich (1998) discussed spatial and temporal variability of stable isotope composition of precipitation in Southeast Asia; their results identified five different air masses controlling the meteorological and pluviometric regime of this area. Tian et al. (2007) identified the northward maximum extent of the southwest monsoon over the Tibetan Plateau with stable isotopes in precipitation. Sengupta and Sarkar (2006) suggested that the precipitation in New Delhi can be attributed to an admixture of Bay of Bengal and Arabian Sea vapor sources, as well as an evapotranspiration effect. The contribution from the Arabian Sea vapor source was estimated at ~20%.
All of the studies mentioned above were based on monthly, weekly, and event-based isotope data. This approach could be considered valid if there were fixed source areas for precipitation and if the physical conditions during the condensation and precipitation in the source areas do not vary temporally, but meteorological regimes (e.g., moisture generation and airmass transport) are diverse and even unfixed for a single weather event, so these assumptions are not usually satisfied. Therefore, isotope-based precipitation research should consider short-term data intervals. Recent studies based on short-term sampling (typically 5–30-min intervals) within single weather events have shown variations of stable isotopic composition in precipitation (Celle-Jeanton et al. 2004; Coplen et al. 2008; Munksgaard et al. 2012). Munksgaard et al. (2012) recorded extreme variations in δ18O (from −8.7‰ to −19.6‰) and δD (from −54‰ to −140‰) within a single 4-h period in Australia. The δ18O and δD of precipitation should be decreased because of the rainout effect, and the break (an increase or a rapid drop) of the isotopic evolution indicates precipitation generation changes. For example, variations in δ18O and δD of precipitation during 24–25 January 2012 in Australia indicated a transition from rain generated in southeasterly oceanic air masses to rain derived from northeasterly air masses (Munksgaard et al. 2012). Different air masses were also discerned by isotopes by Celle-Jeanton et al. (2004) and Coplen et al. (2008). However, the different air masses have not yet been identified and quantified with isotopes in detail.
The present study characterizes and models isotopic evolution during a record-breaking rainstorm in Beijing on 21 July 2012; it was the city’s heaviest rainfall in six decades, delivering 460 mm over the course of 18 h. Diverse isotopic values for two different air masses were calculated with a Rayleigh depletion model and an isotopic mixing model. Combined with δ18O of precipitation at nearby Global Network of Isotopes in Precipitation (GNIP) stations, southwest trajectory in the initial rain stage and a mixture with southeast trajectory in the later stage were identified with a transient rain in between when the two trajectories merged. These results are comparable to meteorological studies (Chen et al. 2012; Grumm 2012; Huang et al. 2014; Sun et al. 2012; Wang et al. 2014; Yu 2012; Zhang et al. 2013) and indicate that using isotopic mixing models to determine the relative contribution of different air masses to precipitation is more reliable than the previously discussed meteorological trajectories method. The results have wide implications for isotope hydrology and isotope climatology–climate change studies.
2. Setting and methodology
a. Sampling site description
The city of Beijing sits at the far northern edge of the North China Plain. The city is bounded to the west by the Taihang Shan and to the north by the Yan Shan and to the east is the Bo Hai, ~150 km at the nearest point. Our sampling at Shihua Cave (39.78°N, 115.93°E) is in the southwest of the Beijing municipality, some 40 km south-southwest of the city center (Fig. 1). Beijing has a typical continental monsoon climate, with four distinct seasons; cool and wet winters are dominated by the polar air mass, while hot and rainy summers are influenced by the East Asian summer monsoon. The annual precipitation is 574 mm, of which some 70% occurs during the summer season of June–September. Average annual temperature is around 12°C (Fig. 2).
Two adjacent GNIP stations, at Taiyuan and Tianjin, were selected to represent the isotopic characteristics of precipitation originating in the southwest and southeast, respectively. The locations of Taiyuan and Tianjin relative to Beijing are shown in Fig. 1 and Table 1. Precipitation amount and temperature data for these stations (1986–2001) were obtained from the China Meteorological Data Sharing Service System (CMDSSS) in agreement with isotopic data from GNIP. Although the climatic features of Taiyuan and Tianjin stations are similar to Beijing, the monthly variances of the isotopic composition of precipitation at the two stations are distinct. Higher δ18O values for June and July were found at Taiyuan station, while more depleted δ18O values were recorded for June and July at the Tianjin station (Fig. 2).
b. Sampling and analytical procedure
A rain collector composed of a polyethylene bottle and a funnel was placed outside of Shihua Cave station, with a ping-pong ball placed at the funnel mouth to prevent evaporation during rainfall. Composite precipitation was sampled monthly as a regularly scheduled collection for isotopic measurement. To measure the immediate response of cave drip water to precipitation during one rainfall event, a collection date was planned for 21 July 2012, based on weather forecast. Fortuitously, our collection station at Shihua Cave was located very near the center point of the record-breaking rainfall event (Fig. 3), receiving 399.8 mm of precipitation over the 18 monitoring hours, which we determined to be valuable data toward understanding of isotopic evolution during a heavy rainfall event. Rainwater samples were collected throughout the rainfall event in 2-L beakers placed on the ground. A total of 30 samples were collected at intervals varying from 10 min to 2 h, with the more frequent samples taken during the initial stage. Samples were immediately sealed in 10-mL brown vitreous bottles, fully filled to avoid evaporation. Total precipitation and air temperature were recorded automatically by an RG3-M data-logging rain gauge, with manufacturer-reported equipment errors for precipitation amount and air temperature of 0.2 mm and 0.47°C, respectively. The samples were transported to the laboratory and stored in refrigeration at 4°C before analysis.
Stable isotopes were analyzed with a Picarro L2130-i laser absorption water isotope spectrometer in the Key Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (Tao et al. 2013). Results are reported as δ18O [δ18O = (Rsample/Rstandard − 1) × 1000] with the standard of Vienna Standard Mean Ocean Water (VSMOW), where Rsample and Rstandard represent 18O/16O ratio for the sample and the reference, respectively. The analytical precision was 0.1‰ for δ18O.
To deduce the probable moisture trajectories of the study area, we generated backward trajectories using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT), which is a development of the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (Draxler and Rolph 2003). Moisture trajectories were mapped at 500, 1000, and 1500 m, each for the 24 h prior to reaching the destination. Figure 4 shows the moisture trajectories at 0400 and 1600 local time (LT; UTC + 8 h) 21 July 2012, showing the change trajectories approaching the study area. The model indicates that moisture at 0400 LT was from the southwest, while moisture at 1600 LT changed to the southeast. This is in agreement with the conclusions of the National Centers for Environmental Prediction (NCEP) analysis, Doppler weather radar, and satellite data (Chen et al. 2012; Grumm 2012; Huang et al. 2014; Wang et al. 2014; Yu 2012), which also determined that the air mass was from the southwest during the initial stage and later changed trajectory from the coast.
The rainfall at Shihua Cave station began at 0937 LT 21 July and ended at 0216 LT 22 July 2012. The stable isotope composition changed remarkably over time, defining four successive stages (Table 2, Fig. 5). Stage 1 lasted from 0937 to 1240 LT 21 July and shows a descending isotopic trend. The temperature during this period drops from 25.6° to 22.4°C, reflecting the progressive adiabatic condensation of atmospheric vapor. The maximum depletion (−10.1‰ for δ18O) was reached at the point of maximum rain intensity (41 mm half-hourly precipitation between 1200 and 1230 LT), which also coincided with the maximum cooling attained as the convection cell reached its peak. Stage 2 was a 2.5-h weakening stage with an increasing trend for isotopes. This is attributed to the arrival of a new air mass, which was also confirmed by multivariate data from meteorological arrays (Grumm 2012; Wang et al. 2014; Yu 2012). Precipitation was low during stage 2, with a rainfall of 28.0 mm, while the temperature remained constant. Stage 3, from 1500 to 2100 LT, showed a descending isotopic trend, with a fast initial drop of about 8‰ for δ18O. Stage 4 held a stationary value, with little precipitation, ending at 0216 LT 22 July.
a. Meteoric water line
Remarkable variations in δ18O (from −6.6‰ to −16.9‰) and δD (from −48.5‰ to −125.6‰) were recorded during this extreme rainfall (Fig. 3). The amount-weighted mean values are −11.5‰ for δ18O and −82.8‰ for δD. If only a cumulative amount-weighted average of δ18O and δD had been calculated, the dynamics of isotopic variations would have been missed (Coplen et al. 2008). Isotopic variations were also reported in single rainfall events in other studies. For example, a δ18O range from −3‰ to −11‰ was observed during rainfall in southeastern Australia (Barras and Simmonds 2008), and a much larger δ18O change, from −8.7‰ to −19.6‰, was reported within a 4-h rainfall in northeastern Australia (Munksgaard et al. 2012).
A significant linear correlation (r2 = 0.99) exists between δ18O and δ2H (Fig. 6). The rapid shifts in δ18O and δD within this rainfall occurred along a local meteoric water line (LMWL). Both the slope (7.8 ± 0.1) and intercept (6.2 ± 1.4) of the linear regression line (RL) show the isotopic characteristics of precipitation in an arid region, which is similar to the local meteoric water line for China (Zheng et al. 1983). It is noteworthy that not only event-based samples plot on the local meteoric water line, but also samples during a rainfall event. This indicates that the storm did not suffer an obvious kinetic effect during and after the rainfall. Samples during stage 1 are located slightly below the regression line, indicating some subcloud evaporation during this stage.
b. Isotopic modeling of rain fraction
We used a numerical Rayleigh distillation model to describe the progressive evolution of the heavy isotope composition of precipitation. When an air mass migrates from its source area to higher latitudes over continents, it cools and loses its water vapor along the way in the form of precipitation; this process is known as rainout. According to a Rayleigh distillation model, at any point along the rainout trajectory the isotopic composition of the residual fraction of vapor mass can be calculated using the following equation:
where and are δ18O values of the residual fraction of vapor and the initial vapor, respectively; is the isotope enrichment factor of 18O for equilibrium water vapor exchange, only dependent on temperature T (K; Majoube 1971); and f is the residual vapor fraction. The rain produced by the vapor mass at any given f is
Using Eqs. (1) and (2), we calculated the residual vapor fractions for samples of stage 1 and combined stages 3 and 4. The term is 9.5‰ under the average temperature of 22.8°C. The correlation between δ18O and the residual vapor fraction f with time as the storm proceeds for stage 1 and combined stages 3 and 4 is evident in Fig. 7a, indicating that rainout is the main reason for the strong depletions in precipitation.
The residual vapor fraction can also be calculated by dividing total water vapor by the amount of precipitation. The values for precipitation are taken directly from the monitoring results recorded by the RG3-M, while the total water vapor of stages 1, 3, and 4 were calculated from the Rayleigh distillation equation (260.0 mm in stage 1 and 525.4 mm in stages 3 and 4; Table 3). The results of the residual vapor fraction calculated with different methods are comparable (Fig. 7b), suggesting that it was the rainout effect that dominated the isotopic evolution of precipitation in both stage 1 and combined stages 3 and 4.
c. Vapor trajectories
The isotopic observation has confirmed the meteorological assessment that identified two distinct air masses. The isotopic data of adjacent GNIP stations (IAEA 2006) are used to represent the isotopic characteristics of precipitation formed by southeast (Tianjin) and southwest (Taiyuan) vapor trajectory. The amount-weighted means of monthly δ18O values for June and July were calculated for comparison.
During stage 1, at 0955 LT 21 July, the precipitation has an initial δ18O of −6.8‰, which falls between the mean values for June (−6.5‰) and July (−7.0‰) at Taiyuan station, suggesting that the vapor trajectory is from the southwest.
To obtain the initial value of southeast vapor trajectory, isotopic binary mixing (Clark and Fritz 1997) was used. At the end of stage 1, at 1240 LT 21 July, the rain had a δ18O of −10.1‰ with 71.0% (184.6 mm) of total water vapor remaining (Table 3). This remaining water vapor was mixed with the external water vapor. The 28.0 mm of precipitation that fell during stage 2 is ignored because of the relatively small precipitation amount, and the remaining vapor from the southwest is assumed to be still in the system. At the beginning of stage 3, the water vapor amount formed by external water vapor will equate to about 525.4 − 184.6 = 340.8 mm. The rain formed from this new mixture of water vapor had a δ18O isotopic composition of −8.7‰ at 1500 LT 21 July, with a total water vapor amount of 525.4 mm. Based on isotopic methods, the proportion of mixing for a given sample relates directly to its position on the mixing line:
where , , and are the δ18O values of a given sample (1 and 2, respectively). The terms P1 and P2 represent the vapor amount for sample 1 and 2, respectively.
With this simple isotopic binary mixing algebra, we can get the initial isotopic composition of this southeast trajectory:
If the remaining vapor from the southwest leaves the system, then the initial precipitation of stage 3 (−8.7‰) will represent the value of vapor from the southeast trajectory. Then, the precipitation δ18O value for the southeast trajectory should be between −7.9‰ and ~−8.7‰. This value is similar to that of the Tianjin station in June (−8.2‰ for δ18O) and July (−8.9‰ for δ18O), suggesting that the vapor was likely from the southeast.
Dansgaard (1964) proposed deuterium excess d first, which is always used as a diagnostic tool to provide information about moisture sources (Fröhlich et al. 2002; Li et al. 2014). In this study, deuterium excess provided further evidence on moisture sources. Deuterium excess was 2.0‰ at the beginning of this extreme rainfall event, indicating a surge of drier air moving from farther inland (Fig. 8a). An increasing trend from 2.0‰ to 10.4‰ for deuterium excess is produced in stage 1, which is attributed to the inverse correlation between δ18O and deuterium excess when the slope of the meteoric water line is lower than 8 (Fig. 8b):
During stage 2, deuterium excess remains around 10‰, suggesting back-building as the source of warm moist air, which was probably extended from offshore areas westward over Beijing. Deuterium excess values of stage 3 remain constant compared to the extreme δ18O variations, indicating the continuous moisture transportation from offshore areas. When the precipitation became sparse during stage 4, an inverse correlation between δ18O and deuterium excess was observed.
The contribution of two moisture sources can also be obtained from Rayleigh distillation and an isotopic binary mixing model. Table 4 shows the results of precipitation amount contributed by different water vapor trajectories. In stage 1, water vapor is predominantly from the southwest trajectory. During stages 3 and 4, if the new air mass from the southeast trajectory is added without missing any remaining moisture (184.6 mm), then the contribution of southwest moisture is
This represents a maximum possible value. The southwest moisture contributes ~0%–35%. At the same time, the southeast moisture contributes ~65%–100%. These results are comparable to the results calculated with HYSPLIT (Wang et al. 2014). Their results also showed that most precipitation seems to be produced by air parcels residing over the eastern sea during this storm.
Isotopic monitoring and modeling of the extreme rainfall event in Beijing on 21 July 2012 have revealed new insights into the formation of the event with respect to the relative contribution of different moisture sources. The rainfall event was divided into four stages based on meteorological and isotopic arrays.
The isotopic composition was reconstructed using the Rayleigh depletion model, which demonstrates that the rainout effect is the dominant process during the rainfall event, except during stage 2, which suggests the incursion of a new moisture source. Rayleigh distillation model indicates that 29% of the initial moisture was precipitated in stage 1, and 56% of moisture was precipitated in stages 3 and 4.
Using a mixing model based on isotopic mass balance, two moisture sources were identified for the rainfall event, comprising a southwest trajectory and a southeast trajectory. The initial isotopic compositions were found similar to long-term monthly average of relevant GNIP stations. The comparison of deuterium excess further confirmed the isotopic evidence. Mixing calculations show that ~65%–100% of the moisture came from the southeast during the storm’s peak rainfall.
This study is supported by the National Natural Science Foundation of China (Grant 41030103 and 41402237), the PhD Student Innovation Program (kjdb2012004), and the China Postdoctoral Science Foundation Funded Project (2013M540138). Many thanks to Ted Crook for his kind help in polishing the English. We also thank two anonymous reviewers for their constructive comments and suggestions that have helped to improve the original manuscript.