Water isotopic variability in response to mesoscale convective system over the tropical ocean



[1] In tropical regions, the empirical negative relationship between the isotopic content of precipitation and rainfall amount, known as the “amount effect”, has been used as a rationale for paleohydroclimate reconstruction from isotope records. However, there is still no comprehensive physical explanation for this empirical effect. Here we reconsider the well-known amount effect using newly available isotope data for both surface water vapor and precipitation obtained from shipboard observations. In this study, we hypothesized that stratiform rainfall associated with mesoscale convective systems (MCSs) is a key process in reducing tropical water isotopic concentrations and tested this hypothesis with an idealized MCS model. Our conceptual model reasonably accounted for several observed features and indicated that isotopic reductions in tropical oceanic regions reflect a precipitating system's change. Relatively high isotope ratios corresponded to disorganized convection. On the other hand, MCSs were characterized by lower isotope ratios with increasing stratiform area. In addition, the amplitude of this isotopic depletion was related to the scale of the precipitation system. The lowest isotopic ratios were observed during the passage of large-scale disturbances—corresponding to the Madden-Julian oscillation convective envelope—in which MCSs are embedded. This means that the frequent appearance of MCSs results in further decreases in the isotopic ratio of surface vapor and precipitation. From this, we conclude that the amount effect can be interpreted as the development of a precipitation system from an isolated convection cell into a large-scale system containing several MCSs and that past large-scale convective activity can be reconstructed from isotope records.

1 Introduction

[2] One of the best ways to unveil past changes in tropical water cycles is to use oxygen- and hydrogen-isotope records, which can be obtained from various paleoarchives (e.g., ice core, stalagmites, tree ring cellulose, leaf waxes, etc.) [e.g., Thompson et al., 2003; Wang et al., 2001; McCarroll and Loader, 2004; Sachse et al., 2012]. Isotopic shifts in such records reflect shifts in meteoric water, and it is well known that the isotopic contents of tropical precipitation decrease as the amount of local precipitation increases under present-day climatic conditions (the amount effect [Dansgaard, 1964]). Based on this relationship, isotope records obtained from low-latitude regions are often used for paleohydroclimate reconstruction [e.g., Cruz et al., 2005, 2009; Niedermeyer et al., 2010; Tierney et al., 2008; Thompson et al., 2000; Partin et al., 2007; Sano et al., 2012]. However, this effect is not universal over the tropics and subtropics. Robust relationships were only observed at monthly or longer time scales at tropical marine island stations [Rozanski et al., 1993]. Over land, although a clear negative trend with local precipitation was seen at each station, the slope of the regression line differed considerably among stations. Therefore, plots obtained from land stations are scattered, as is seen in Figure 1. In addition, recent observations have shown that short-term (daily or event-based) isotopic variations are independent of local precipitation intensity, even at marine island stations [Kurita et al., 2009]. These complexities raise the question of whether precipitation amount is a key factor controlling isotopic variation in tropical precipitation. In tropical regions, most precipitation falls from convective clouds and is accompanied by intense vertical water transport. Surface vapor is lifted by convection, forms condensed water in the updraft, and is removed as precipitation. When precipitation falls through an unsaturated atmosphere, reevaporation of precipitation occurs and induces a downdraft, which transports moisture from the free troposphere to the lower atmosphere. For many years, the amount effect has been explained by the condensation-related mechanism. Because isotopically enriched water vapor is preferably removed as condensation, the isotopic content of the remaining vapor becomes lighter. Thus, the more rainfall that forms from a given convective air mass, the more depleted the isotopic content of the vapor and subsequent precipitation will be [e.g., Cole et al., 1999; Yoshimura et al., 2003; Vuille et al., 2003]. However, this cannot fully explain all aspects of the amount effect. Recently, a counterexplanation was proposed by Risi et al. [2008a]. Using a single column model including Emanuel convective parameterization, they reasonably reproduced the amount effect derived from marine island stations and proposed the downdraft recycling hypothesis, in which water vapor that is transported by downdraft to the lower atmosphere is reused for precipitation. Because lighter isotopes evaporate preferentially, the reevaporation of falling rainfall results in a lowering of the isotopic ratio in atmospheric vapor, then this isotopically depleted vapor is injected through downdraft mass flux into the subcloud layer. The more intense the convection, the stronger the convective downdrafts, and the more depleted the isotopic content in the subcloud layer vapor feeding the convective system by convective flux [Risi et al., 2008a]. This hypothesis can reasonably explain why short-term isotopic variations do not correspond to instantaneous precipitation amount. Risi et al. [2008a] investigated the timescale of the amount effect using the same model and showed that the isotopic content of precipitation reflects the integrated history of convective activity over 4 days. This concept can explain more aspects of this effect than before and seems to be plausible. However, because the single column model was driven by an idealized framework, more investigation is required to verify that this downdraft recycling theory can be applied to real convective systems.

Figure 1.

Relationship between δ18O in precipitation and amount of local precipitation in the lower latitude region (30°N–S). Data are long-term mean monthly average values for precipitation and were obtained from the Global Network for Isotope in Precipitation (GNIP) database. Black circles (gray crosses) indicate stations on the marine island (land) in the GNIP database.

[3] In the real world, there is a wide variety of convective systems, ranging from isolated convection to organized convective systems in size and from shallow cumulus to deep cumulonimbus in hight [e.g., Lopéz, 1976; Houze and Cheng, 1977; Mapes and Houze, 1993; Rickenbach and Rutledge, 1998; Masunaga et al., 2005; Masunaga and Kummerow, 2006]. In the tropics, individual storms mostly organize upscale into single cloud systems, referred to as cloud clusters and mesoscale convective system (MCSs). MCSs are ubiquitous features over the tropics and account for a large portion of tropical rainfall. MCSs are defined as an organized ensemble of convective elements, which have a contiguous precipitation area of approximately 100 km or more in at least one dimension [Houze, 1993]. The typical structure of MCSs, shown schematically in Figure 2, is characterized by a leading convective line followed by much broader stratiform rain areas [e.g., Zipser, 1969, 1977; Houze, 1977; Houze et al., 1989]. The convective region consists of an intense, vertically extending convective core with strong heavy local precipitation, while the stratiform region displays widespread gentle lifting (mesoscale ascent) and results in weak precipitation intensity. However, the area covered by stratiform rainfall is much larger than that occupied by convection cells. Shumacher and Houze [2003] reported that more than 70% of the rain area in tropical regions is covered by stratiform precipitation using Tropical Rainfall Measurement Mission (TRMM) precipitation data. The stratiform region has a cloud base in the middle troposphere, and, due to the cooling of midtropospheric environment air through the melting and evaporation of precipitation, cold air descent occurs below the stratiform cloud base (mesoscale subsidence) (see Houze Jr. [2004] for a detailed review). This mesoscale subsidence may contribute to the injection of isotopically depleted vapor into the subcloud layer (hereafter this recycling process associated with MCSs is referred to as vapor recycling). In addition, MCSs are often embedded in large-scale systems and are accompanied by tropical waves that move parallel to the equator [Takayabu, 1994; Wheeler and Kiladis, 1999]. Well-known examples include westward propagating synoptic disturbances within the intertropical convective zone (hereafter referred to as ITCZ) [Chang, 1970]. The largest component of this is the Madden-Julian oscillation (MJO) [Madden and Julian, 1971], which is the most prominent mode of intraseasonal variability in the tropics. The MJO is characterized by an eastward moving super cloud clusters (SCCs) with a horizontal scale of several thousand kilometers; an SCC consists of several MCSs [e.g., Nakazawa, 1988]. From this, the intensity and duration of mesoscale subsidence may vary in relation to the characteristics of a large-scale system, such as the spatial scale and life cycle of that system.

Figure 2.

Schematic diagram of the idealized mesoscale system water budget. The water budget parameters Cu, Ecd, Ece, CA, Cmu, Emd, Eme, PRc, and PRm are determined referring to Gamache and Houze [1983].

[4] Isotopic response to organized convective systems has been investigated using ground-based observation data. Lawrence et al. [2004] first pointed out that the lowest isotopic signals in water vapor occur in or downwind of organized convective clouds. Recently, several studies reported that isotopic variation in precipitation corresponds to the large-scale convective activity associated with intraseasonal variation [Risi et al., 2008b; Vimeux et al., 2011; Kurita et al., 2011; Tremoy et al., 2012; Moerman et al., 2013]. In particular, Kurita et al. [2011] highlighted that the isotopic depletions observed during the convectively active phase are linked to MCS activity. They found that the lowest isotopic peaks in surface vapor correspond with maximum values in stratiform rainfall areas in MCSs and showed a clear relationship between lower isotopic values and increases in stratiform rain areas. This relationship supports the vapor recycling theory. However, this was only reported in relation to an MJO event. To ensure that this relationship is robust throughout the tropics, in addition to the data used in Kurita et al. [2011], we use here newly available isotope data for both surface water vapor and precipitation gathered over tropical and subtropical oceans. First, we examine the spatial and temporal features of isotopic values in surface vapor over the ocean, and then we test the vapor recycling theory with another MCS.

[5] Furthermore, in this study, to reinforce the conclusions obtained from hydrogen or oxygen isotopes, we also used the d-excess (d-value) tracer, which was defined by Dansgaard [1964] as d=δD-8δ18O where delta δ is inline image (R is the isotopic ratio (HDO/H2O and inline image) and V-SMOW is Vienna Standard Mean Ocean Water). The d-value in vapor is recognized as a tracer that indicates moisture origin because it primarily reflects environmental conditions, such as relative humidity, wind speed, and the sea surface temperature where the evaporation took place [Merlivat and Jouzel, 1979]. However, recently, Kurita et al. [2011] proposed that d-values in the tropical atmosphere can be used as a tag of subsidence. By using an isotope-enabled general circulation model (GCM), they showed that subsiding air from stratiform rainfall regions is characterized by having lower δD values and higher d-excess. Although the vertical profile of d-values has not been reported in previous observational studies, calculated d-values from isotopic profiles reported by He and Smith [1999] support their conclusion. The d-excess values at the top of a boundary layer are significantly higher than those at the surface. In addition, recent aircraft observation of water vapor isotopologue measurements in the upper troposphere showed an increase in d-values with altitude [Sayres et al., 2010], and a similar increasing trend in d-values was simulated in a single column model [Bony et al., 2008]. Furthermore, although it does not provide direct evidence, the fact that observed d-values in tropical precipitation increased with altitude [Gonfiantini et al., 2001] is consistent with this theory. From these indications, we regard d-value in surface vapor to be a good indicator for the evaluation of vertical mixing between boundary layer and free troposphere in the tropical region.

2 Data and Methods

2.1 Isotope Observations Over Ocean Areas

[6] A survey of water isotopologues in atmospheric water vapor and precipitation over oceans is being conducted by Independent Administrative Institution, Japan Agency for Marine-Earth Science and Technology (JAMSTEC). Surface water vapor and precipitation samples have been continuously collected aboard the Research Vessel (R/V) Mirai throughout all of its research cruises. This sampling program was initiated in September 2006 [Kurita et al., 2011] and became operational after April 2008. Approximately 20 expeditions were carried out between April 2008 and March 2012, and the route of travel differed for each expedition. A composite map summarizing the sampling locations is shown in Figure 3a. Although the sampling locations were concentrated in the western Pacific region, our data cover a wide latitudinal range, from the Arctic Ocean to the southern Pacific Ocean (50°S).

Figure 3.

(a) Cruise tracks of R/V Mirai observations used in this study. (b) Mean observed δD in surface vapor, shown on grids of 4°×4° without any interpolation or smoothing.

[7] Air sampling was conducted using a conventional cold trap method [Kurita, 2011; Kurita et al., 2011]. Ambient air was drawn from an air intake at a height of 20 m above the sea surface through 9 mm OD nylon tubing and a glass trap that was submerged in an ethanol bath, which was thermoelectrically cooled to −100°C. The flow rate was set at 1.5∼3.0 L min−1, depending on the vapor concentration, and water vapor was collected in the cold trap over a 6∼24 h period. After sample collection, water in the trap was thawed and then stored in 6 mL capacity glass bottles. For precipitation sampling, a rain collector with a 100 cm2catchment area was set on the upper deck and rain water samples were collected after each rainfall event. Through this observational program, more than 2000 water vapor and 650 precipitation samples were collected up to March 2012. Isotopic analysis of collected water was performed at the JAMSTEC laboratory using cavity ring-down spectroscopy isotopic water analysis (model L1102-i; Picarro Inc., Sunnyvale, CA, USA) with a CTC Analytics autosampler (model HTC-PAL; Leap Technologies, Carrboro, NC, USA). Measurement precision, including internal and external variations, was better than ±0.2‰ for δ18O and ±2‰ for δD. Thus, analytical uncertainty in d-excess for our measurements is better than 2.1‰. During the research cruises, surface meteorological data (air temperature, humidity, wind speed and direction, etc.) were measured at the same mast height; sea surface salinity and surface temperature data are also available.

2.2 Satellite Data

[8] A Tropical Rainfall Measuring Mission (TRMM)-adjusted merged infrared (IR) precipitation product 3B42 version 6 (3B42) was used to estimate area-averaged precipitation at R/V Mirai locations. The 3B42 data set is a 3 h averaged surface rain rate product that has a resolution of 0.25°×0.25° over the latitudinal band 50°N–S. This product is derived from IR-based geosynchronous satellite rainfall estimates that are calibrated using a merged-microwave precipitation estimate, which consists of TRMM, a Special Sensor Microwave Imager, an Advanced Microwave Scanning Radiometer, and an Advanced Microwave Sounding Unit. Infrared (IR) brightness temperature data from Meteosat-5 (Indian Ocean) and MTSAT-1R (western Pacific) were also used to provide data on large-scale meteorological conditions during the field campaigns; these were undertaken at a fixed site. Both satellites collect images at hourly time intervals and the image resolution is 0.25° for Meteosat-7 and 0.04° for MTSAT-1R.

2.3 Modeling Isotopic Content

2.3.1 Water Budget

[9] A simple conceptual model based on water budget analysis was used to understand and quantify isotopic response to the vapor recycling associated with MCSs over the tropical ocean. The vertical water cycle of an idealized MCS is schematically shown in Figure 2. In convective region, surface vapor is lifted by convection and forms condensed water (Cu) in the updrafts. Much of condensate falls out of the updraft and either evaporates in the convective downdraft (Ecd) or precipitates to the surface as heavy rainfall (PRc). The rest of the condensate is detrained from the convective cells (as the sum of cloud condensate and vapor) and then either evaporates into the environment (Ece) or is carried into the mesoscale stratiform cloud adjacent to the convection (CA). In the stratiform region, mesoscale ascent above the melting layer produces ice particles (Cmu), which can either fall from the stratiform cloud base or be transferred by horizontal flow out of the cloud (Eme). The ice particles that fall from stratiform clouds melt into liquid drops below the melting layer and then either evaporate (Emd) into unsaturated air in the mesoscale subsidence or reach the surface as relatively light rainfall (PRm). Following Houze et al. [1980], the budget for total condensed water can be expressed as for the convective region

display math(1)

and the other for stratiform region

display math(2)

where Wc(Wm) is total condensed water in the convective (stratiform) region. Assuming a steady state, these water budget terms have been determined from direct measurements based on Doppler radar observations of tropical squall lines [Leary and Houze, 1980; Gamache and Houze, 1983; Chong and Hauser, 1989]. By using these water budget parameters as boundary fluxes, we simulate the isotopic evolution associated with the vertical water cycle of MCSs.

2.3.2 The Five-Box Model of MCSs

[10] A five-box model used in this study is illustrated in Figure 4a. This model consists of two boxes for the convective region, two boxes for the stratiform region, and one box for the surface layer. The boxes are connected by a flow of water vapor and the steady state field is computed using fixed moisture flux and typical meteorological conditions based on observation. In addition, to simplify the model, we assume a well-mixed column of water, so that the isotopic content of vapor and condensate is homogeneous in the box. Generally, the transportation of heavy isotope species is treated like “normal water” as long as no phase changes occur. Isotopic change happens during condensation in convective updrafts, in stratiform clouds, and during the reevaporation of falling rainfall into the unsaturated atmosphere. In addition, isotopic exchange also takes place between small raindrops in stratiform rainfall and the surrounding air before reaching the surface.

Figure 4.

(a) Schematic diagram of the five-box model of the idealized mesoscale system shown in Figure 2. Arrows represent water budget of vapor (white) and condensate (black). Symbols are defined in section 2.3.2 of the text. (b) Same as Figure 4a, except the variables have been replaced with their values as used in this study. Values of components in water budget corresponds to Gamache and Houze [1983]. These numbers are expressed as fraction of the convective rainfall (PRc). The atmospheric boundary condition (PW and Tw) derived from observed meteorological data is also shown.

[11] Over the tropical ocean, surface vapor can be considered to be an admixture between surface evaporation flux and downward mass flux across the boundary layer. Therefore, the isotopic ratio of surface vapor feeding the convective system inline image can be written using the recycling ratio (γ)

display math(3)

where Rrecycling and inline image are the isotopic ratio of downward moving air across the boundary layer (recycling vapor) and surface evaporation flux.

[12] Box 1 (the lower box in the convective region) is characterized by strong convective updraft. We assumed that all condensed water (Cu) is formed in box 1 and is removed as precipitation (PRc+Ecd). The isotopic ratios in water vapor (Rq1) and condensate (Rc1) in box 1 are provided by Rayleigh distillation and water mass conservation

display math(4)
display math(5)

where inline image represents the equilibrium fractionation factor between liquid and vapor at the water vapor weighted mean temperature in updraft, inline image is the isotopic ratio in the surface layer (box 5), and fc is the fraction of condensed water in the updraft (fc=Cu/PW1, PW1 represents the saturated column water content in box 1). The subscript number is the number of the box. The isotopic ratio of surface convective rainfall (inline image) can be expressed as

display math(6)

where inline imageis the isotopic ratio of evaporation flux from convective rain.

[13] After the removal of precipitation, the rest of the condensate (together with the vapor) in the convective updraft is transported into box 2 (the upper box in the convective region) and is detrained from convective clouds. Using mass balance, the isotopic ratio of total water detrained from convective updraft (Rqt2) can be written as

display math(7)

CA reported by Gamache and Houze [1983] corresponds to cloud water flux. Thus, in this study, water vapor flux to produce condensate in stratiform (Cmu) region is added to original CA value in Gamache and Houze [1983].

[14] As for the stratiform region, two boxes are divided above and below the stratiform cloud. In box 3 (the upper box in the stratiform region), water advected from the convective region (CA) is mixed with background water vapor (PW3), then mesoscale ascent produces ice particles. The isotopic ratios in water vapor (Rq3) and ice particles (Rc3) are shown by

display math(8)
display math(9)
display math(10)

where inline image(inline image) represents the isotopic ratio of water vapor in box 3 before (after) mixing with advected water from the convective region, fm is the fraction of ice particles formed in box3 (fm=Cmu/(CA+PW3)), and inline imagerepresents the kinetic fractionation factor between ice crystals and water vapor at the water vapor weighted mean temperature in the stratiform cloud. In this model, we assume that ice formation occurs in a supersaturated environment and calculate this kinetic fractionation factor following Jouzel and Merlivat [1984]. In box 4 (the lower box in the stratiform region), melting of solid precipitation occurs and melted raindrops then reevaporate into the unsaturated air. Although no isotopic change is associated with the melting of ice because ice particles melts totally, fractionation, including kinetic effects, takes place during the reevaporation of falling rainfall. In addition, we consider the diffusive exchange between falling drops and the surrounding vapor. Recent modeling studies highlighted the importance of this diffusive exchange process [Risi et al., 2008a; Barras and Simmonds, 2009; Yoshimura et al., 2010]. These isotopic changes can be estimated using Stewart's formula [Stewart, 1975]. Following Stewart's theory, we calculate the isotopic ratio of reevaporation flux (inline image). In box 4, the isotopic content of water vapor (inline image) and precipitation (inline image) is expressed as

display math(11)
display math(12)

where inline imagerepresents the isotopic ratio of water vapor prior to mixing with evaporated vapor. According to Jouzel [1986], small raindrops can reach isotopic equilibrium with the surrounding vapor within a few seconds. Thus, we further assume that most of the stratiform rainfall reaches isotopic equilibrium with the surrounding vapor in box 4.

display math(13)
display math(14)
display math(15)

where Feq is the portion of equilibrated raindrops (Feq=0.95), inline imagerepresents the equilibrium fractionation factor at the water vapor weighted mean temperature in box 4, and ζrepresents the fraction of stratiform rain of the total water content. Rq4 and Rc4 correspond to the isotopic ratio of the recycling vapor (Rrecycling) and surface stratiform precipitation (inline image).

2.3.3 Experimental Design

[15] The simulation is initialized using plausible isotopic values of water vapor in each box then water vapor circulation in the box model is driven by imposing water fluxes described in the water budget section until all initial conditions within the cycle converge into the steady state. The values of water budget terms (PRc,Cu, Ecd, Ece, Eme, and Eme) are determined based on Gamache and Houze [1983]. Although Gamache and Houze [1983] assumed total rainfall from the MCSs (14.1×1011 kg) with 50% coming from the convective region (7.2×1011 kg, which corresponds to average rainfall rate of PRc= 5.3 mm h−1) and 50% from the stratiform region (6.9×1011 kg, which corresponds to average rainfall rate of PRm=2.3 mm h−1), we assumed that the term of PRm is variable to control the stratiform rainfall fraction (SRF), holding it between 10% and 50% of the total rainfall. The water budget parameters used in this study are shown in Figure 4b. The numbers in Figure 4b are expressed as a fraction of the convective rainfall (PRc), which in our case is 5.3 mm h−1.

[16] The atmospheric boundary condition (average condensation temperature, water content in each box) is derived from observed meteorological data gathered during the field campaign. The average temperature and relative humidity profiles are calculated from upper air observation data collected over the tropical ocean. We then calculated humidity weighted average temperature (Tw) and total column water (PW) in each box. The isotopic value of the surface evaporation flux is assumed to be the highest value in surface marine vapor in our data set. These atmospheric parameters used in this five-box model are also shown in Figure 4b.

[17] The main purpose of this modeling is to examine the isotopic variation in response to vapor recycling associated with MCS. Thus, each run was done by changing the recycling ratio γ from 0 to 0.9.

2.4 Rainfall-Type Classification

[18] Rainfall-type classification is usually based on the horizontal distribution of Doppler radar reflectivity (rainfall intensity). A C-band Doppler radar was mounted on the R/V Mirai, and convective/stratiform rainfall separation was undertaken following the work of Steiner et al. [1995] when radar observation was conducted. However, because radar observations were not carried out during every cruise, we looked for the substitute for stratiform rainfall area using surface rainfall intensity data derived from TRMM 3B42. As described in the introduction, although stratiform precipitation is weaker than convective precipitation, the area of a stratiform rainfall region is much larger than that occupied by convective cells. This means that an increase in stratiform area coverage results in more intense area-averaged precipitation. Figure 5a shows the relationship between area-averaged precipitation rates and area coverage of stratiform precipitation obtained from Doppler radar observations during the Mirai Indian Ocean Cruise for the Study of the MJO-Convection Onset (MISMO) field campaign, which was carried out over the equatorial Indian Ocean in 2006 [Yoneyama et al., 2008]. As expected, the results show a statistically robust (P<0.01) increase in area-averaged rainfall intensity associated with a stratiform precipitation area. Therefore, we use area-averaged rainfall derived from the TRMM 3B42 as a substitute for Doppler radar-derived stratiform rainfall area coverage. The area-averaged rainfall intensity (inline image) was calculated using the following equation:

display math(16)

where Pi,j is the surface rainfall rate over pixel i,jwith areas Ai,j and Atotal being the total area of the domain (1.5°×1.5°), which corresponds to the total coverage area of Doppler radar observations that were taken by the R/V Mirai. Finally, we evaluated the calculated inline imageby comparing area-averaged rainfall derived from Doppler radar data (3 h average data). As shown in Figure 5b, most data plotted along the 1:1 line, which indicates that variability in calculated inline image is highly linked to that of the stratiform rainfall area.

Figure 5.

(a) The Doppler radar-derived area-averaged precipitation rate (inline image) as a function of area coverage of stratiform precipitation during the MISMO field campaign. A linear regression analysis is plotted as solid line. (b) Comparison of area-averaged precipitation derived from Doppler radar observation and from TRMM 3B42 during the MISMO field campaign. The 3 h averaged precipitation data in the total coverage area of radar was used for comparison.

3 Results

3.1 Spatial Distribution of HDO

[19] The spatial distribution of δD observed in this study is shown in Figure 3b. Despite sparse data coverage, the observed distribution showed a clear decrease in δD with latitude from the subtropics to high latitude regions. This echoes the well-known “latitude effect” that is found in observations of δD in precipitation. In contrast, in the tropics, δD values vary considerably and lower δD values are often observed along the equator. This spatial feature can be seen in the global map of near surface HDO in atmospheric vapor derived from the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography [Frankenberg et al., 2009]. Does this latitudinal variation result from the isotopic shift of evaporated moisture from the ocean surface? To answer this question, we calculated the δD values in marine vapor based on Merlivat and Jouzel [1979] with assumption of closure. This assumption means that all the marine vapor consists of locally evaporated moisture. Figure 6a shows a comparison of the observations with a calculated zonal mean latitudinal δD gradient. Although a gradually decreasing trend with latitude is consistent between observation and calculation, observed zonal mean δD values are much more depleted than calculated values. Interestingly, the calculated δD values correspond to the maximum observed values. This suggests that marine moisture is not simply constituted from local evaporation flux but that other sources with low δD values significantly contribute to surface marine vapor.

Figure 6.

Latitudinal distribution of (a) δD and (b) d-excess values obtained from R/V Mirai cruises. Means are indicated by blue solid lines inside the boxes; the upper and lower limits indicate the 75th and 25th percentiles, respectively. Maximum and minimum values are also shown as horizontal lines. The red line represents the mean of the calculated δD/d-excess values in marine vapor using the closure assumption model based on Merlivat and Jouzel [1979]. Dashed red lines are maximum and minimum of simulated values.

[20] In the midlatitude region (>30°N), meridional transport of air masses is intense, owing to the activities of transient eddies. A transient eddy may transport not only heat but also moisture from higher latitudes. Thus, the contribution of advected moisture with low δD values from high latitudes may result in the lower δD in observed water vapor. In addition, vertical mixing between the boundary layer and free troposphere may also contribute to δD depletion in surface vapor. In the subtropical region (15°–30°, N–S), variability in δD is smaller than at midlatitudes. Subtropical high pressure systems persist throughout the year, and thus, this region is characterized by low rainfall and high evaporation flux from the ocean. The fact that dominant δD values in subtropical vapor are close to the calculated values with closure assumption suggests that surface evaporation is the dominant source of surface moisture (Figure 7). The slightly depleted δD values observed may reflect the influence of vertical mixing between the boundary layer and free troposphere due to shallow convection. On the other hand, in the tropical region, the mean δD value was largely lower than in the subtropics. Although Figure 7 shows that the dominant δD value is similar to that in the subtropics, there are much lower δD values in the tropics. We will discuss the influence of MCSs on these lower isotopic values in the next section.

Figure 7.

Histograms of δD values in surface vapor in the (a) subtropical and (b) tropical ocean areas where sea surface temperature (SST) exceeded 25°C. The red line represents the calculated average δD values in tropical marine vapor (SST>25°C) using the closure assumption model based on Merlivat and Jouzel [1979].

3.2 The d-excess

[21] The zonally averaged latitudinal d-excess pattern is shown in Figure 6b. Although the mean values of d-excess calculated using Merlivat and Jouzel [1979] with closure assumption do not show a clear latitudinal trend, the observed data display a gradual increasing trend with latitude. This implies that transient eddies may transport moisture having high d-excess values from high latitude regions. It is known that high d-values appear whenever evaporation occurs under conditions with a large humidity deficit and a strong temperature contrast between surface air temperature and sea surface temperature [e.g., Gat et al., 2003]. In the northwestern Pacific, especially in winter, the monsoon brings a cold and dry continental air mass to the relatively warmer ocean; thus, evaporation occurs under such conditions. Therefore, the contribution of advected moisture from near the continent must result in a higher d-excess in surface vapor. Except for in the midlatitude region, the mean of calculated d-values is a little bit smaller than that of the observation value. This is also consistent with the influence of free tropospheric air mixing into the boundary layer, because the d-excess values at the top of the boundary layer are relatively higher than those at the surface [He and Smith, 1999].

4 Isotopic Response to MCSs

[22] Over the tropical ocean, there is a large spread in the δD values of surface vapor. These δD values show two clear relationships in association with vertical water cycle. First, as shown in Figure 8, δD variation in surface vapor (δDvap) is closely related with variation in precipitation (δDpre) (R2=0.899 and P<0.01). This positive linear correlation suggests that surface vapor feeds the convective updraft and condensates to form precipitation. The second is that lower δDvap values are accompanied by higher d-values. In Figure 9a, all of the observations from subtropical and tropical regions (30°N–S) are plotted on an δD-d-excess diagram. When δD values gradually decrease, a clear trend between lower δD values and higher d-values is evident. However, in the case of high δD values, d-values are distributed without a clear trend. As was mentioned in the previous section, when evaporation occurs under conditions with a high saturation deficit, evaporation flux shows high d-values. Thus, samples observed under dry conditions, when the relative humidity of the surface air was less than 70%, displayed relatively high d-values (see Figure 9a). It is noteworthy that the highest d-values were observed in the subtropical western Pacific under humid conditions during the winter. Except for these data, all other δD values were distributed along the negative linear relationship that was obtained from the MISMO campaign [Kurita et al., 2011]. The GCM output, Kurita et al. [2011] explained this negative relationship as follows: subsidence in the stratiform region transports water vapor with low isotopic content and high d-excess to the boundary layer.

Figure 8.

Relationship between 12 h average δD values in surface vapor and those in precipitation derived from R/V Mirai observations (cross). The black solid line is a regression line. Dashed red lines indicate simulated evolutions of δD in precipitation by changing water vapor recycling ratio. The number in red circles represents recycling ratio. Several model simulations were done with different stratiform rainfall fraction (SRF) from 20% to 50%.

Figure 9.

(a) δD-d-excess plots for surface water vapor derived from R/V Mirai observations. Colors are used to denote average relative humidity during the moisture trapping period. The regression line (R2=0.322) was calculated using data collected only during the MISMO campaign (open circle). Black squares represent data obtained from the Western Pacific during the winter (December–February). (b) Sensitivity of δD-d-excess relationship in surface vapor to relative humidity (RH) when reevaporation of precipitation occur (solid blue line). The experiment which doubled precipitation (PRc=10mm/h) is shown as dashed sky blue lines. The experiment run that no isotopic exchange occurs between falling rainfall and surrounding air is represented as dashed red lines. In these experiments, stratiform rainfall fraction (SRF) is assumed at 30%.

[23] Here we examine whether these linear relationships can be explained by isotopic evolution through water vapor recycling associated with MCSs. By using a simple conceptual model of MCSs, variation in δDpre values with increases in the contribution of recycled moisture are computed and then plotted against δDvap values in Figure 8. Similar to the observation, the δDpre values varied linearly with δDvap; however, the slope of this trend is much higher than the observed value (0.72). We tuned each model parameter and found that the slope value is basically insensitive to water budget parameters and is only sensitive to the convective/stratiform rainfall ratio of the total rainfall. In this model, because air mass that undergoes convective rainfall forms stratiform rainfall, δDpre in stratiform rainfall is much lower than that in convective rainfall. The reevaporation of rainfall tends to result in heavy isotope enrichment in stratiform rainfall, however this effect is weakened by reequilibrium process between rainfall and environmental air. Thus, when the stratiform rainfall fraction (SRF) to total rainfall increases, the line moves parallel to the lower δDpre value. Figure 8 shows the gradual decrease in δD associated with increasing SRF, and this is consistent with the result obtained from MISMO campaign [Kurita et al., 2011]. Relatively high δD values corresponded to the disorganized convection. On the other hand, the lowest δD values are plotted around the relationship in case SRF is 50%, which corresponds to the MCS reported by Gamache and Houze [1983]. According to Shumacher and Houze [2003], SRF derived from satellite observation mainly varies from 20% to 50% over the ocean, and thus this, although indirect, is in agreement with the range of δD variation in Figure 8. From these, the observed small slope value can be interpreted as that precipitating system's change from disorganized convection to an organized convective system. In addition, Figure 8 also suggests that recycling ratio gradually increase with SRF.

[24] A comparison of the δD-d-excess relationship in surface vapor between the observation and model output is shown in Figure 9b. In our model, the trend of lower δDvap values with increases in d-excess is primarily governed by two parameters that control isotopic kinetic effect associated with the formation of ice and the postcondensation effect of falling stratiform rainfall. The ice crystals formed in the upper box are labeled by high d-excess values and that signature is then carried to the vapor in the lower box through rain reevaporation and isotopic exchange processes. In the lower box, the relative importance of reevaporation and diffusive exchange depends on relative humidity (RH). When the RH tends toward 100%, rain reevaporation does not occur but diffusive exchange takes place and isotopic equilibrium is reached between rainfall and environmental air. Generally, diffusive exchange tends to be heavy isotope enrichment in precipitation and depletion in surrounding vapor. D-excess in the vapor is the weighted mean between the compositions of the original vapor and rainfall. Contrarily, the reevaporation of rainfall dominates for low RH. Due to the differing diffusive behavior of each water isotope, reevaporation of rainfall results in lower (higher) δD and higher (lower) d-excess in environmental air (precipitation). This process enhances the sensitivity of RH to δD-d-excess relationships. When RH is set at 100%, the simulated increasing trend in d-excess with a decrease in δD depends on the kinetic process during ice formation. The increase in d-excess gradually weakens with decreasing RH. In addition, this tendency is more clear as stratiform precipitation increases (sky blue dashed lines). Interestingly, this RH sensitivity opposes the above explanation based on the relative contribution of reevaporation and diffusive exchange. In this model, we assumed that 95% of rainfall reaches isotopic equilibrium before arriving at the surface and that this reequilibration process plays an important role in this contrary behavior. Without the isotope equilibrium process, the d-excess in surface vapor increases more rapidly (red broken line). Reevaporation of rainfall results in lower d-excess in precipitation and reequilibration with these raindrops acts to decrease the d-excess in environmental air. In addition, this redistribution between the vapor and raindrops depends on the fraction of stratiform rain of the total water content. Although our conceptual model is too simple to discuss the δD-d-excess relationship, the fact that this simple conceptual model can reasonably account for observed trend with appropriate parameters indicates that our model grasps the essence. Thus, we conclude that stratiform rainfall associated with MCSs is a key to the decreasing isotopic values over the tropical ocean.

5 Isotopic Footprint in Marine Vapor

[25] Finally, we discuss the affected area in relation to isotopic values in surface water vapor. Using MISMO field campaign data, Kurita et al. [2011] showed a clear relationship between δD in surface vapor and the stratiform rainfall fraction within the Doppler radar observation coverage area (80,000 km2). This reminds us that isotopic variation may be linked to mesoscale precipitation systems. To ensure this, δD values observed over the tropical ocean from 2006 to 2010 were plotted against average regional precipitation in Figure 10. The statistically robust (P<0.01) gradual decrease in δD that was associated with increasing average regional precipitation can be seen in each expedition. However, these linear trends are not identical to the results obtained from the MISMO campaign. Although differences in slope among expeditions are relatively small, the intercept derived from the MISMO campaign is about 10‰ lower than from the others. Except for the MISMO figures, intercepts are close to the δD in locally evaporated moisture (Figure 7). Because the large decreases in δD during the MISMO campaign correspond with the MJO, isotopic values in surface vapor are not controlled by the mesoscale but rather, they reflect more large-scale phenomena. In Figure 11, temporal δD variations in surface vapor are illustrated, together with the large-scale precipitation field and cloud top temperature found during the MISMO and Palau-2010 field campaigns, which were conducted at a fixed site. Both field campaigns displayed common features. The lowest δD peaks corresponded to the passage of synoptic-scale precipitation systems, which propagate westward following the easterly flow. In the case of one MJO event, during which the record minimum of δD was observed, several streaks of westward moving cloud shields were observed going away from the leading eastward moving precipitation (EP) system, and it was at the intersection of these clouds that the lower isotopic values occurred [Kurita et al., 2011]. In contrast, δD values increased and were near the isotopic values of evaporated sea surface water when fine weather continued for several days in the windward region before those systems arrived (labeled X and Y in Figure 11). These results highlight the fact that past convective activity in the windward region can influence observed isotopic values.

Figure 10.

Relationship between δD values in surface vapor and regional-mean rainfall intensity (inline image) derived from TRMM 3B42 in the tropical region (10°N–S). Each color represents a different R/V Mirai expedition. The least squares linear regression line for each expedition is shown using the same color as the markers.

Figure 11.

(top) Time series of 6/12 h average δD values and radar echo area coverage within which reflectivity exceeded 15 dBZ for convective-type (orange) and stratiform-type (green) clouds at a height of 3 km during the MISMO field campaign (left) and the PALAU-2010 campaign (right). The characters A–I for MISMO and 1–12 for Palau-2010 indicate the maximum peaks in the stratiform rainfall fraction with δD minima in surface vapor. The EP with yellow bar indicates the passage of a leading edge of an eastward propagating precipitation system. The characters XY correspond to fine weather that was sustained in the upstream region. (bottom) Time-longitude section of cloud top temperature (°C) averaged over 5°N–S intervals for MISMO and 2.5°N–7.5°N intervals for the Palau-2010 campaign, and δD values in surface vapor at the R/V Mirai location. The contour lines indicate surface rainfall intensity (> 0.4 mm/hour) derived from TRMM 3B42 data, averaged over the same latitudinal bands as cloud temperature. Solid red lines indicate identified EP systems. Dashed black lines indicate the average propagation speed of cloud shield (14 m/s). The characters A–I (1–12) refer to the same points as in the top panel.

[26] To verify this finding, we used accumulated regional-scale rainfall (Pacc). The Pacc is defined as an integrated indicator of past regional-scale precipitation that influenced the observed surface air and was computed along the surface air mass trajectory (see Appendix). The relationship between the δD in surface vapor and Pacc over 4 days is displayed in (Figure 12). The optimal accumulation time (τacc) was chosen as 4 days because it showed the highest correlation between the δD in surface vapor and Pacc over τacc days. Figure 12 shows the clear decreasing trend in δD with Pacc and the duration of rainfall. When surface air does not encounter rainfall events, δD reaches its maximum values. On the other hand, the lowest δD values occur when convective activity is sustained for several days. Over the tropical ocean, this corresponds to the passage of large-scale disturbances by embedded MCSs (supercluster). Figure 13 schematically illustrates the mechanism of this large decrease in δD in surface vapor which accompanies the passage of a westward moving supercluster. Generally, because propagation speed of a supercluster is much faster than surface wind, multiple MCSs in a supercluster can influence the δD value of surface vapor. Each MCS moves isotopically depleted vapor to the surface via mesoscale subsidence, so that the δD value in surface vapor gradually decreases with increase in the contribution of recycled vapor, while active convection is sustained and reaches its lowest value downwind from the supercluster. After a heavy rainfall ends, surface heat flux from the ocean supplies moisture with relatively enriched isotopic ratios to the atmosphere and thus surface isotopic values increase gradually. This explanation is consistent with record minimum isotopic values occurring during an MJO, which is the largest convective envelope in the tropics. From these results, we can conclude that isotopic values in surface vapor reflect not only mesoscale MCS activity but also large-scale convective activity.

Figure 12.

Relationship between δD values in surface vapor and accumulated regional-mean rainfall amount (Pacc) over a 4 day back trajectory of the air mass in the tropical region (10°N–S). Colors are used to denote the total number of hours of precipitation over the 4 days. Inset: Variation in the correlation coefficient (R2) for the δD-Pacc relationship with variation in the period of accumulation. The highest R2 value was observed with 4 days of accumulated rainfall.

Figure 13.

Schematic illustrations of temporal δD evolutions in surface water vapor and precipitation that accompanied the passage of westward propagating large-scale disturbance.

6 Summary and Discussion

[27] A global survey of water isotopologues in marine vapor showed that enriched isotopic values, which correspond to locally evaporated moisture from the surface of the ocean, are dominant in the warm pool region; however, largely depleted values occasionally occur in association with the MCSs. These systems contain a large region of stratiform rainfall and in this stratiform region evaporative cooling induces mesoscale subsidence and injects isotopically depleted vapor into the lower troposphere. MCSs are often embedded in large-scale disturbances and the frequent appearance of MCSs results in further decreases in the isotopic value of surface vapor during the passage of those systems. The amplitude of this isotopic depletion is closely related to the scale of a precipitation system, so that record minimum isotopic values correspond to a convectively active period during an MJO. In addition, because surface vapor is a dominant source for tropical rainfall, this isotopic variation mirrors that of precipitation. These results indicate that MCS activity must play a key role in the amount effect.

[28] Over the tropical ocean, it is well known that large-scale disturbances occur in subseasonal time scales of 30 to 60 days. In a convectively inactive region/period, precipitation falls from local convection systems (e.g., isolated convective storms); thus, locally evaporated moisture is the dominant source of boundary layer moisture and convective rainfall. The isotopic content of precipitation is characterized by high values. On the other hand, in a convectively active region/period, the majority of rainfall falls from propagating synoptic-scale disturbances, which consist of several MCSs. Successive rainfall from MCSs results in increased precipitation over a region and decreased isotopic values. Therefore, clear negative isotope-precipitation relationships are found in monthly mean or longer time scale data, which correspond with the intraseasonal timescale, even though similar relationships are weak in daily average data.

[29] Over land, similar intraseasonal isotopic variations associated with large-scale convective activity have already been reported in South America [Vimeux et al., 2011], in Africa [Risi et al., 2008b; Tremoy et al., 2012], and in South Asia [Moerman et al., 2013]. These intraseasonal isotopic variations are dominant during the monsoon season. Monsoons transport moist air from oceanic areas, and the atmosphere can achieve favorable conditions for the development of organized convections. Thus, organized precipitation systems containing MCSs appear frequently during the monsoon season [Laing and Fritsch, 1993; Miller and Fritsch, 1991]. This indicates that isotopic variations in response to MCS activity are ubiquitous features all over the subtropics/tropics. However, this conclusion raises a question: Why is the amount effect at sites over land more ambiguous than that over the ocean? Over land, the diurnal cycle of convective activity in rainfall is more prominent and thus strong, but short-lived convection accounts for a significant proportion of annual rainfall. Because convective rainfall is shutdown every day, the time that is available for growth in stratiform regions is limited and convection systems over land do not contain the large stratiform regions that are observed over the ocean, unless some external force (such as monsoon flow) overcomes the diurnal cycle [Shumacher and Houze, 2003]. These factors may contribute to distorting the amount effect.

[30] From these results, we can conclude that isotopic variation in tropical/subtropical regions is not directly controlled by rainfall amount but depends on large-scale convective activity containing several MCSs. This may lead to development of the interpretation of the isotope proxy. In the study of paleoclimate, based on the amount effect, oxygen- and hydrogen-isotope records are commonly interpreted as the proxy for past precipitation amounts. However, our findings propose that isotope proxies can be used to reconstruct past large-scale convective activity. A shift to lower values in the isotope records of paleoarchives can be interpreted not as an increase in local precipitation but as an increase in the genesis scale or frequency of large-scale precipitation systems. The fact that record minimum isotopic values were encountered during the active MJO phase leads us to expect that there is some possibility that past MJO activity can be reconstructed from isotope proxies. Progress in the understanding of present-day isotopic variation may lead to new interpretations of existing isotope proxies.

Appendix A

[31] Accumulated regional-scale rainfall (Pacc), which is integrated area-averaged precipitation along the air mass trajectory, was computed using the following equation

display math(A1)

where the subscript n indicates the time step of trajectory analysis and inline image is the area-averaged rainfall rate at point x,yon the trajectory at time t. Air mass trajectories are calculated 7 days backward in time using cross-calibrated, multiplatform (CCMP) ocean surface wind data with a resolution of 0.25°×0.25° [Atlas et al., 2011]. A trajectory was launched from the location of the ship at surface level when the sample was collected. Because the ship moved during the period of sample collection, the midposition of the ship during the sampling period was used as the launching point. The horizontal location of a parcel is provided in its implicit form by

display math(A2)
display math(A3)

where the subscript n indicates the time step and the zonal u and meridional vwinds are provided over a time step of 1 h. CCMP data for 6 h periods was interpolated to hourly time steps.

[32] A problem in determining accurate Pacc has been data availability. Because of the usage of high-resolution satellite precipitation (0.25°×0.25°) product, reliable wind data with the same resolution is required for this analysis. However, wind field data obtained from global reanalysis data are too coarse to apply to this study. In addition, because the number of upper air sounding stations is limited over the oceanic region, these data may include large uncertainties. Thus, in this study, we only used horizontal surface wind data derived by satellite (CCMP); thus, vertical air mass motion cannot be considered. This means that we assumed that grid average upward flow is much slower than horizontal advection.


[33] This research was supported by a grant for environmental research projects from the Sumitomo Foundation, Japan. We thank Klaus Froehlich, Camille Risi, and anonymous reviewers for useful comments regarding this paper and Maureen Macneill for editing the original text. We thank the technical staff of Global Ocean Development Inc. for their support during Arctic Ocean observations. The TRMM 3B42 data were obtained from the National Aeronautics and Space Administration Goddard Earth Sciences Data and Information Service Center. The MTSAT-1R data were acquired from the Center for Environmental Remote Sensing (CEReS), Chiba University. The cross-calibrated multiplatform (CCMP) ocean surface wind product was obtained from the NASA Jet Propulsion Laboratory (JPL), Physical Oceanography Distributed Active Archive Center (PO.DAAC).