Hydrogen and oxygen isotopic compositions of water vapor over the Bay of Bengal during monsoon



[1] δDv and δ18Ov of ~70 water vapor samples collected at 6 and 25 m above sea level over the Bay of Bengal (BoB) during July–August 2012 are reported. This helps characterize the isotopic signature of monsoon vapor. No significant vertical variation is observed in δDv, δ18Ov, or deuterium excess (defined as δD–8δ18O); δDv and δ18Ov are significantly correlated (r = 0.92) at each height; the deuterium excess values do not, because the variation of δDv and δ18Ov relative to their uncertainties is much larger than that of the latter. The temporal variations of δDv and δ18Ov correlate well with air temperature rather than sea surface temperature. The control of normalized humidity on deuterium excess is less prominent. While the distribution of water vapor isotopologues over the BoB is primarily determined by the ocean surface conditions, they are significantly altered by laterally advected vapor from rain en route during the monsoon.

1 Introduction

[2] Stable water isotopologues, H218O and HDO, are widely used as tracers of the hydrological cycle [e.g., Gat, 1996] and paleomonsoon [e.g., Ramesh et al., 2010]. The relative mass difference of H218O and H216O (as well as HDO and H216O) causes isotopic fractionation during phase change of water, including evaporation from open water bodies, condensation in clouds, and below-cloud reevaporation while rain falls through the undersaturated atmosphere. Isotopic fractionation during large-scale condensation is believed to be an equilibrium process, obeying Rayleigh isotopic fractionation [Yurtsever and Gat, 1981]. However, isotopic exchange processes in clouds formed due to intense convective activity are more complex [Bony et al., 2008; Bolot et al., 2012]. For treating isotopic fractionation during evaporation, a model proposed by Craig and Gordon [Craig and Gordon, 1965, hereinafter referred to as the C-G model] and later modified by Merlivat and Jouzel [Merlivat and Jouzel, 1979, hereinafter referred to as MJ79] is widely used. The relative abundance of the heavier isotope is usually represented as a delta value; δ = (Rsample/RVSMOW–1) × 103‰, where Rsample is the ratio of the abundances of the heavier to lighter isotope of interest (e.g., D/H or 18O/16O) in the sample, and RVSMOW is its value in the standard, here the Vienna Standard Mean Ocean Water [Yurtsever and Gat, 1981]. According to MJ79, the isotopic ratio of the evaporating flux is estimated by the equation

display math(1)

where RE, RL, and RA are, respectively, the isotopic ratios of the emanating vapor, ocean surface water, and preexisting atmospheric vapor. α, k, and h are, respectively, the equilibrium and kinetic fractionation factors between coexisting vapor and liquid water and the relative humidity at 10 m above sea level, normalized to saturation at the sea surface temperature (SST). Equation ((1)) is widely used in isotope-enabled general circulation models to parameterize the isotopic composition of the evaporated vapor [e.g., Joussaume et al., 1984; Yoshimura et al., 2008]. MJ79 predicts a relationship between ocean surface relative humidity and deuterium excess (dv = δDv 8δ18Ov) by applying a global closure assumption (i.e., <RE > = < RA > for global average values) in the equation. Though this leads to a systematic bias in predicting the vapor isotopic values [Jouzel et al., 1996], the relationship between h and d-excess is indeed observed in the marine vapor over the Southern Ocean [Uemura et al., 2008]. This relation is yet to be verified in the tropical oceans, a major moisture source for the global hydrological cycle. Here we attempt to understand the influence of boundary layer processes and ocean surface conditions on the stable hydrogen and oxygen isotopic compositions (δ18Ov and δDv) of atmospheric vapor over the Bay of Bengal (BoB). These are the first measurements being reported from the Bay of Bengal, one of the important sources for South Asian monsoon rain.

2 Sampling and Isotopic Measurements

[3] Water vapor samples were collected cryogenically using a glass trap maintained below –80°C with an ethanol liquid nitrogen bath (IAEA protocol, http://www-naweb.iaea.org/napc/ih/documents/miba/water_vapor_protocol.pdf), at locations shown in Figure 1. To ensure efficient trapping, the flow speed of air through the trap was maintained at 500 mL/min, sampling duration ~3 h to trap water vapor >2 mL in liquid form. The efficiency of the trap was checked by connecting an extra cold trap to the outlet of original trap, and no significant condensate was found in it. Forty-two samples of vapor were collected from the top mast (25 m above sea level) of R/V Sagar Kanya (cruise #SK-296). We carried out another simultaneous collection at 6 m height, but the number of samples at this height is less (n = 28) due to limited availability of liquid nitrogen for trapping. Rain water samples (n = 15, this number is limited by the occurrence of rain) and sea surface water samples (n = 42) were also collected, and all samples were analyzed for δD and δ18O using a Thermo Delta-V-Plus isotope ratio mass spectrometer. The H2O-CO2 equilibrium method was adopted for δ18O measurements [Epstein and Mayeda, 1953], while the H2O-H2 equilibration in the presence of a platinum catalyst was used for δD measurements (for more details, see Srivastava et al. [2010]). The precision values of the measurements are 1‰ for δD and 0.1‰ for δ18O, and the propagated uncertainty in the estimation of d is 1.3‰.

Figure 1.

Seventy-two hour air mass back trajectories from the sampling locations with (a) relative humidity, (b) air parcel height along the trajectory, and (c) specific humidity (in grams/kilograms) along the trajectories for the 14 samples collected prior to 19 July. (d) Same as Figure 1c but for the rest of the 28 samples. Black dots in the diagrams represent sampling locations.

3 Results and Discussion

3.1 Air Parcel Trajectory Analysis

[4] For all days of water vapor collection, 72 h air mass back trajectory analysis was done using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [Draxler and Rolph, 2003]. The GDAS (Global Data Assimilation System) data [Kanamitsu, 1989] is used as input to this model, which generates three-dimensional back trajectories of air parcels and meteorological variables such as rainfall, pressure, potential temperature, temperature, and relative humidity along the trajectory.

[5] Due to the strong westerly monsoon winds, air parcels traveled from the Arabian Sea to the BoB during the sampling period. While crossing the southern peninsular India, air parcels dehydrated due to condensation over the Western Ghats. Model-derived heights of air parcels show a descending motion of air to the east of the Western Ghats and mixing with hotter, drier air in the lower troposphere (the so-called “rain shadow effect”). This caused the low relative humidity over peninsular India, and the parcel gained moisture after entering the BoB (see Figures 1a and 1b). A major amount of the moisture collected by us thus seems to have originated from the BoB alone. There is no reported descending motion of air over the BoB during this active monsoon season, implying that there is little chance of mixing of air above the atmospheric boundary layer with the advected air parcel coming from over peninsular India.

3.2 Variations of δD and δ18O of Vapor and Meteorological Parameters

[6] Figure 2 shows the time series of stable isotopic composition of water vapor and the associated meteorological parameters. δ18Ov values at 25 m varied from 9.8‰ to 14.1‰ (mean ± standard deviation, −11.4 ±0.9‰), while δDv varied from 69.1‰ to 97.2‰ (−78.2 ±5.7‰); the dv varied from 6.9 to 19.4‰ (12.6 ±3.0‰); at 6 m, δ18Ov varied from −10.0 to −13.6‰ (−10.9 ± 0.8‰), δDv varied from −68.3‰ to −94.0‰ (−74.6 ±5.4‰), and dv varied from 5.7‰ to 16.4‰ (12.7 ± 2.4‰). Stable isotopic compositions of sea surface water varied from −2.5‰ to 0‰ for δ18Os (−0.4 ±0.4‰) and from −8.9‰ to 5.2‰ for δDs (−3.1 ±2.1‰), and ds varied from 3.7‰ to 11.1‰ (6.3 ±1.2‰). The isotopic compositions of vapor and dv at 6 and 25 m did not show any statistically significant difference. However, the linear correlation coefficients (r) between δDv at 6 and 25 m (r = 0.78) and that between δ18Ov values (r = 0.63) are significant. However, the dv values between these heights are uncorrelated (r = −0.13). The reason is that the isotopic ratios show much larger range of variation relative to their respective experimental errors (a factor of 36–43 for δ18Ov and 26–28 in δDv while only 8–10 in dv) than dv.

Figure 2.

Time series of (a) δ18Ov, (b) δDv, and (c) dv values of marine vapor at 6 m (open symbols) and 25 m height (filled symbols) above sea level. (d) Air temperature and sea surface temperature. (e) Relative humidity and normalized humidity h. (f) Latent heat flux. (g) Atmospheric pressure at 25 m above sea level.

[7] The 3 hourly averaged wind (i.e., averaged over the sample collection time) was less than 10 m/s up to 21 July, and it increased up to 20 m/s later (22 July to 6 August 2012). The measured latent heat flux also shows a similar trend with a linear correlation coefficient r of 0.78 (P < 0.01) with the wind speed. However, such a trend is observed in neither δDv nor δ18Ov; rather, they show a correlation [r = −0.66 (P < 0.01) for δ18Ov and r = 0.68 (P < 0.01) for dv] with the latent heat flux and with the wind speed [r = −0.53 (P < 0.01) for δ18Ov and r = 0.47 (P < 0.05) for dv] during the latter sampling period. Two marked dips are observed in δDv and δ18Ov, associated with the presence of monsoon depressions at the collection site, marked by a fall in atmospheric pressure and an increased rainfall (on 20 July and 4 August 2012).

3.3 Influence of Local Ocean Surface Conditions

[8] Figure 2 is useful to infer the relations between isotopic compositions of atmospheric vapor and sea surface meteorological conditions. The expected relation between relative humidity and d (due to diffusive transport) was not observed as SST variations were smaller than those in air temperature; thus, it appears that relative humidity may not be a good indicator here of kinetic fractionation during evaporation. As predicted by MJ79, dv, however, is inversely correlated with normalized humidity h. Twenty-five percent of the variance in dv is explained by the normalized relative humidity [d = (−0.55 ± 0.14)*h + (56 ± 12), r = 0.5, P < 0.01]. Interestingly, the regression slope of −0.55‰/% and intercept 56‰ agree within the cited uncertainties, respectively, with the slope of −0.61‰/% and intercept 55‰ obtained by Uemura et al. [2008] for marine vapor over the Indian sector of the Southern Ocean (35–65°S and 20–115°E) during the Austral summer of 2006 (however, h correlated with relative humidity, i.e., r = 0.9 in their study). These results perhaps point toward the possible existence of a global relation between h and dv as predicted by equation ((1)), although the coefficient of determination here is less than that in the Southern Ocean (r ~ 0.8) [Uemura et al., 2008]. This may be a result of less temporal variability of h over the BoB during the sampling monsoon period or may be due to the weakening of the h-dv relation above h = 80%, as seen in isotope-enabled global models [Uemura et al., 2008]. Atmospheric air temperature also shows a significant positive linear correlation δDv and δ18Ov [r = 0.61 (P < 0.01) for δDv and r = 0.62 (P < 0.01) for δ18Ov], while they are uncorrelated with SST observations. This is likely due to the cooling of surface air during rainfall and associated isotopic equilibration of vapor with falling raindrops. Rain-vapor interactions are again discussed in section 3.4. Laser spectroscopy for continuous vapor sampling [e.g., Kurita et al., 2012] would be more useful to increase the sample throughput to put such observations on a firmer footing.

[9] The factors determining the variations in δ18Ov and dv sampled up to 21 July 2012 and later appear to be quite different. The later samples exhibit better correlations with meteorological parameters such as air temperature, normalized humidity, latent heat flux, and wind speed than the earlier ones. Figures 1c and 1d show that less moisture is advected to the sampling location in the later than the earlier period. As local evaporation contributed more vapor than advection during the later period, this resulted in better correlation coefficients as observed (Table 1). Another reason could be that the spatial variability in the position of the vessel was more restricted for about 10 days in this period.

Table 1. Linear Correlation Coefficients (r) Between Different Parameters for All the Collected Samples (Second Column) and Only Samples Collected After 21 July 2012 (Third Column)a
ParametersAll Samples (n = 42)Post 21 July 2012 (n = 27)
  1. a

    Double star indicates significant r values with P < 0.01, single star indicates P < 0.05, and no star indicates insignificance. As δ18Ov and δDv are significantly correlated, only correlation with δ18Ov is presented.

Normalized humidity with d-excess−0.50**−0.66**
d with latent heat flux0.190.68**
δ18Ov with air temperature0.63**0.80**
δ18Ov with latent heat flux−0.17−0.66**
d with wind speed0.240.47*
δ18Ov with wind speed−0.07−0.53**

3.4 Comparison With the Craig and Gordon Model and Influence of Advected Moisture

[10] Figure 3 shows the comparison of data with predicted values from the C-G model, although the global closure assumption [<RE >=< RA> in equation ((1))] is not applicable on a regional scale. The equilibrium fractionation factor (α) is taken from Majoube [1971a, 1971b], and the kinetic fractionation factor k is calculated using wind-dependent parameterization proposed by MJ79. The model estimates that δ18Ov is closer to observation during nonrainy days and not on rainy days. This is due to exchange and reevaporation from the falling raindrops, not taken into account by the C-G model. In addition, downdrafts during convective rain events can bring vapor with depleted δ values and higher dv values from the boundary layer above to the surface [Knupp and Cotton, 1985; Kurita, 2013]. We infer that it is imperative to account for (a) isotopic exchange between vapor and raindrops [Stewart, 1975], (b) the mixing with the boundary layer vapor laterally advected to the collection site, and (c) vertical mixing during convective downdraft to significantly improve the model prediction.

Figure 3.

(a) Comparison of observed variations in δ18Ov (filled circles) with C-G model results (inverted triangles). The shaded area shows three rain spells that occurred during sampling. The rain rate (open square) plotted in the upper panel is the average rain rate along the 24 h air parcel trajectory. The accumulated rain amount collected during the cruise (vertical bars) and its δ18Or (filled circle) are also shown in bottom panel. (b) The relation between the deviation between the observed δ18Ov (represented as δ18Oobs) from the C-G model result [represented as Δ(δ18O) in the regression line] and the average rain rate along the 24 h back. The R in the regression equation represents the rain rate. (c) Same as Figure 3b but for δDv.

[11] The strong southwesterly winds provide a continuous supply of the moisture to the sampling location, so the isotopic exchange occurring along its trajectory needs to be considered to explain the observed variations in δ18Ov. Evaporation of and isotopic exchange with falling raindrops along the air parcel trajectory may cause advection of depleted vapor to the collection site. The observed negative correlation of r = −0.62 between the observed deviations in δ18Ov (−0.65 for δDv) from the C-G model prediction and the average rain rate along the trajectory for the previous 24 h (HYSPLIT-derived rain rate) prior to sampling (see Figure 3) is consistent with the above hypothesis. Recently, Kurita [2013] have also shown that the rain activity along the air parcel back trajectory significantly depletes the isotopic composition of surface vapor over the tropical oceans. When raindrops evaporate into unsaturated air, vapor relatively depleted in 18O (and D), with higher dv values results, as observed by us during the three spells of rain (shaded area in Figure 2).

4 Conclusion

[12] The δDv and δ18Ov over the BoB are controlled mainly by the ocean surface conditions, while a lateral advection of vapor could alter their values significantly. The relation between deuterium excess and normalized humidity appears to be valid for the tropics, though it is less prominent than over the Southern Ocean, due to smaller variations in humidity during the monsoon. δDv and δ18Ov values correlate well with atmospheric temperature rather than SST. While the mean values of δDv, δ18Ov, and d-excess at 6 and 25 m do not differ significantly, the d-excess values between these heights do not correlate as the isotope ratios do, as the variation in the former is relatively low. During the dry days, the Craig and Gordon model results are closer to the observed δ18O, while during rainy days, δ18O is more depleted as a result of lateral advection of vapor derived from reevaporation from falling raindrops.


[13] We thank the MoES/CTCZ program and the captain crew, and participants of ORV Sagar Kanya cruise (#SK-296) for their assistance. We thank G. S. Bhat, CAOS, IISc, for providing latent heat flux data and Vijayakumar, NIO, Goa, for providing AWS data. We thank ISRO-GBP for the funding.

[14] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.