In Situ Monitoring of Water Isotopic Composition for Vapor and Precipitation Near-Surface Ground in East Asia Subtropical Monsoon Region

Abstract

Hydrogen and oxygen isotopes in atmospheric water vapor (δ_v) and precipitation (δ_p or δ_r) were continuously measured using a laser-based water isotope spectrometer in Guangzhou, southeastern China, from March 2016 to February 2018. The measurements were conducted to investigate the variations in water isotopes in the hydrological cycle under the subtropical monsoon climate. The isotopic composition ranged from −24.4‰ to −11.1‰ for δ¹⁸O in water vapor (δ¹⁸O_v) and from −11.5‰ to 2.3‰ for δ¹⁸O in precipitation (δ¹⁸O_r). The values of δ_v and δ_r were enriched during the dry season and depleted during the wet season, exhibiting systematic seasonal variation. A negative correlation was observed between monthly δ_v and precipitation amount, indicating that the values of δ_v exhibits an ‘amount effect’. However, a corresponding amount effect was not observed in the values of δ_r. The mean difference between δ_v and δ_r was −9.7‰ for δ¹⁸O and −76‰ for δD, suggesting that equilibrium fractionation is the dominant process during precipitation. The local meteoric vapor line (LMVL) for Guangzhou (δD = 6.6δ¹⁸O − 6.4) exhibited a slope similar to that of the equilibrium local meteoric vapor line (ELMVL) but with an intercept difference of 8.6. This difference in intercepts can be attributed to the vertical profile of δ_v. The δD-q (q refers to water vapor concentration) relationship is useful for identifying water vapor sources and tracking isotopic changes during atmospheric transport and precipitation. The local water vapor was found to originate primarily from the mixing of oceanic air masses. Data points falling between the oceanic source mixing line and the Rayleigh curve likely reflect post-condensation processes, such as raindrop re-evaporation or mixing with surrounding ambient vapor. Short periods of heavy precipitation were observed to cause severe depletion in δ_v, resulting in values falling below the Rayleigh curve.

1. Introduction

Stable isotopes of oxygen and hydrogen in water (δ¹⁸O, δD) are well-established tracers for studying the atmospheric water cycle at a range of spatial and temporal scales [1,2,3]. Due to their stability, they are widely utilized not only for identifying water sources, the investigation of surface and groundwater formation conditions and hydrological processes, but also for paleoclimate reconstruction and the validation of global circulation models (GCMs) [4,5,6]. Traditional interpretations of precipitation isotopic ratios rely on the temperature effect, amount effect, altitude/elevation effect and continental effect [1,2]. However, the isotopic composition of atmospheric water vapor (δ_v) contains additional information on evaporation, large-scale mixing and kinetic fractionation that is not captured by precipitation alone.

The transport of atmospheric water vapor is a process with spatiotemporal continuity. Consequently, the analysis of δ_v allows for a more comprehensive investigation of the redistribution and replenishment components within the water cycle, facilitating a specific understanding of the characteristics and history of water vapor across various spatiotemporal scales. Over the past decade, the development of laser spectrometer technology has facilitated rapid advancements in the observation of δ_v from the surface to the troposphere [7]. Isotopic data from water vapor have been introduced to refine GCMs [6], to enhance the understanding of hydrological cycles at different scales (e.g., evapotranspiration partitioning) [8], and to verify and evaluate satellite-based observations of δ_v [9]. The volume of δ_v datasets has increased tremendously over the past decade, and various explanations for the observed isotopic variations have also been proposed. On a seasonal scale, the monthly values of δ¹⁸O and δD in water vapor have been shown to exhibit a unimodal trend [10], and diurnal cycles in δ_v have also been documented [11]. The spatiotemporal variations of δ_v have been found to be influenced by monsoon activity [12], extreme climatic events [13], air mass trajectories [14], tropical depressions [15], and local geographical conditions [16]. Furthermore, diurnal variations in δ_v have been linked to processes such as local evapotranspiration [10] and convection [12].

Research on precipitation isotope ratios (δ_p or δ_r) in the Southeast Asian monsoon region has a long history, notably through the Global Network of Isotopes in Precipitation (GNIP), which was initiated in 1960 by the International Atomic Energy Agency (IAEA) and the World Meteorological Organization (WMO). Through the GNIP program, long-term datasets of δ_r, spanning over 50 years, have been established for locations such as Hong Kong, Diego Garcia Island, and Mumbai. However, few long-term observations of δ_v in the Southeast Asian monsoon region, as well as concurrent measurements of both δ_v and δ_r in its coastal areas, have been published. Nevertheless, some short-term measurements, such as a month-long study of δ_v within a mangrove ecosystem, have been reported [17]. Given that South China is one of the world’s most climatically complex regions, influenced by both the southeast and southwest monsoons, the study of δ_v in this area can provide crucial information for paleoclimatic interpretations derived from stable water isotope records. Therefore, the present study was undertaken in Guangzhou to investigate the δ_v of near-surface water vapor and δ_r. The primary objectives were: (1) to characterize the local isotopic signatures of atmospheric water; (2) to analyze the meteorological variables influencing δ_v; (3) to compare the differences between δ_v and δ_r; (4) to determine the dominant isotopic fractionation process during precipitation; and (5) to assess the utility of water vapor isotopes for identifying moisture sources.

2. Method

2.1. Measurements and Calculations

The monitoring site for δ_v and δ_r was established at the Panyu Campus of Sun Yat-sen University in Guangzhou, China (113.32° E, 23.13° N; elevation ~6 m a.s.l.). The site is characterized by a typical subtropical monsoon climate, with a mean annual temperature of 22.8 °C. The mean annual precipitation over the past three decades is 1797.9 mm (The average meteorological data of Guangzhou from 1974 to 2014 is obtained from the website of China Meteorological Data Service Center (CMDC, http://data.cma.cn/, accessed on 1 January 2019)). The coastal region of Southern China is primarily influenced by three main moisture source pathways: the Indian Ocean, the South China Sea, and the western Pacific Ocean (Figure 1). Four meteorological parameters (air temperature, relative humidity, wind speed, and precipitation) were continuously measured and recorded by an on-site automatic meteorological station. See the supplementary information for more on the sample collection process.

A shielded air inlet was installed on the building’s roof at an elevation of 30 m above ground level. This inlet was connected to an ultra-high precision isotope analyzer (Picarro L2130-i, Picarro Inc., Sunnyvale, CA, USA). Water vapor was continuously sampled and measured at a rate of 34 L/min by means of a vacuum pump. A minimum of 20 s was required to obtain values of δ_v, and to mitigate memory effects, a 30-min average value was chosen to represent hourly vapor isotopes. When the relative humidity exceeded 95%, high-purity Perfluoroalkoxy (PFA) tubing was carefully monitored for condensation. If condensation occurred, the affected section of the tube was immediately heated. Instrument calibration was performed annually using Chinese National Standards, which are traceable to the IAEA scale as detailed in the supplementary material. Furthermore, Working Standards were employed to calibrate the water samplers and verify the instrument’s status. Additional details regarding the calibration process are provided in the supplementary file [18,19,20,21,22,23,24,25].

In total, 239 daily rain samples were collected. These samples were sealed in 50 mL plastic bottles with Parafilm and stored at 4 °C prior to analysis using the ultra-high precision isotope analyzer. The isotope analyzer is capable of measuring up to 100 liquid samples per run, including four standard samples. Typically, isotopic measurements of liquid samples are performed within one week of their collection. However, during periods characterized by infrequent precipitation, the samples are held and analyzed only after a collection period of one month has been completed.

The simultaneous observation period for both water vapor and precipitation spanned 128 days. The isotope ratios of water are presented in parts per thousand, as follows:

$${\delta }^{18}O\left(\delta D\right)=({\displaystyle \frac{R_S}{R_{V-SMOW}}}-1)$$

(1)

where $R_S$ and $R_{V-SMOW}$ denote the ratios of stable isotopes of oxygen or hydrogen (¹⁸O/¹⁶O or D/H) for the sample and Vienna Standard Mean Ocean Water (V-SMOW), respectively. Following processing with the SICalib program, provided by the IAEA and incorporating memory correction, drift correction, calibration, and uncertainty calculation, the precision for liquid water was determined to be 0.04‰ for δ¹⁸O and 0.4‰ for δD. Similarly, the precision for water vapor was 0.30‰/60 s for δ¹⁸O and 1.11‰/60 s for δD. The monthly weighted average of δ_r is calculated using the following equation:

$${\delta }_r={\displaystyle \frac{{\sum }_{i=1}^n{\delta }_{r,i}\times P_i}{{\sum }_{i=1}^n{P}_i}}$$

(2)

where ${\delta }_r$, i and $P_i$ represent the isotope ratio of precipitation and the amount of precipitation on rainy day i, respectively. Here, n signifies the number of rainy days. Precipitation isotopes can include those from both rainwater and snow. For this study, only rainwater samples were collected. Therefore, the subscript label for isotope ratio of precipitation is set to “r” instead of “p”.

Assuming that precipitation adheres to equilibrium isotope fractionation during precipitation, the equilibrium isotope value (δ_e) in water vapor can be estimated from δ_r and the equilibrium fractionation factor (${\alpha }_e$) [26]. Subsequently, an equilibrium local meteoric vapor line (ELMVL) can be established. If the ground elevation is known, the isotope ratios of water vapor and liquid water can be estimated for a given height, such as the cloud base level [27]. The isotope ratio of water vapor (${\delta }_z$) at a specific height (z) can then be calculated using the following formula [7]:

$${\delta }_z=\left[\left\{\left(1+{10}^3{\delta }_g\right)exp\left(-({\alpha }_e-1)(z/z_s)\right)\right\}-1\right]\times 1000$$

(3)

where ${\delta }_g$ represents the surface groundwater isotope ratio of water vapor, ${\alpha }_e$ denotes the equilibrium fractionation factor at the specific temperature, and $z_s$ is the height of the water vapor scale (the average cloud base height).

When the value of δ_v is positively correlated with temperature, δ_v can be described as a function of specific humidity for Rayleigh fractionation. The following equation was used to represent the value of δ_v [7]:

$${\delta }_v={\delta }_0+({\alpha }_e-1)\ln \left(q_v/q_o\right)$$

(4)

In this equation, ${\delta }_0$ and ${\delta }_v$ refer to the initial and final isotope ratios of water vapor, respectively. Similarly, $q_o$ (25 mmol/mol) and $q_v$ are the initial and final specific humidities, respectively. The term ${\alpha }_e$ (at 25 °C) is defined as previously stated.

2.2. Moisture Source Diagnostic

A characterization of the water vapor transport reaching the study site was performed using humidity source diagnostic calculations. These calculations were based on the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, a tool developed jointly by the Air Resources Laboratory (ARL) of the National Oceanic and Atmospheric Administration (NOAA) and Australia’s Melbourne Meteorological Research Institute. Global Data Assimilation System (GDAS) meteorological data, with a horizontal resolution of 1° × 1°, were utilized as input. HYSPLIT provides specific meteorological variables as outputs, including hourly pressure, altitude, temperature, and specific humidity along each trajectory.

Trajectories of air masses arriving at the observation point at the same daily time (18:00 UTC) were calculated separately. To identify the vapor transport sources contributing to terminal precipitation, an analysis of the hourly position and humidity variation in air masses was conducted. This analysis focused on parcels with humidity <0 in the high-altitude sink area at a starting point of 1500 m. Subsequently, eligible moisture trajectories were selected for clustering analysis. The total number of air mass trajectories clustered over a three-month period was used as a proxy for one season.

3. Results

3.1. Meteorological Measurements

Air temperature ranged from 3.3 °C to 37.0 °C, with an average of 21.9 °C (

Table 1. Monthly mean values of temperature, relative humidity, precipitation amount and the isotopes values of vapor (δ_v), precipitation (δ_r) and equilibrium vapor (δ_e).

						Vapor				Precipitation			Equilibrium Vapor
Year	Month	P (mm)	T (°C)	RH %	V (m/s)	δ18Ov (‰)	δDv (‰)	dv (‰)	H2O (mmol/mol)	δ18Or (‰)	δDr (‰)	dr (‰)	δ18Oe (‰)	δDe (‰)	de (‰)
2016	Mar	253.8	14.89	89.06	1.72	−13.2	−90	15.2	20.17	−2.1	−7	9.5	−12.6	−88	12.6
2016	Apr	272.6	22.42	89.37	1.56	−12.1	−86	9.9	29.64	−1.3	−5	5.7	−11.2	−82	7.7
2016	May	297.9	25.01	86.35	1.87	−13.5	−94	13.8	32.32	−2.5	−13	6.7	−12.2	−87	10.4
2016	Jun	520.5	27.52	90.20	1.40	−14.8	−105	13.4	36.01	−5.3	−35	7.7	−16.0	−113	14.7
2016	Jul	301.4	28.19	86.90	1.82	−15.2	−106	15.4	36.33	−5.5	−37	6.9	−15.2	−109	12.6
2016	Aug	425.8	26.98	90.55	1.54	−14.5	−101	14.7	34.67	−2.0	−13	3.6	−16.1	−115	14.2
2016	Sept	211	26.01	85.97	1.84	−18.6	−130	18.2	31.17	−5.1	−34	6.9	−17.5	−126	13.9
2016	Oct	141.5	23.68	85.84	2.48	−16.8	−118	16.5	29.57	−2.9	−18	4.6	−14.2	−105	8.4
2016	Nov	62	17.51	88.33	2.25	−16.2	−114	16.4	22.17	−0.7	−5	1.1	−16.5	−122	10.0
2016	Dec	3.4	13.99	76.19	2.44	−14.1	−98	14.5	19.70	−2.2	−9	9.3	−14.0	−98	13.7
2017	Jan	13.4	13.44	90.41	2.19	−14.4	−102	13.3	18.77
2017	Feb	26.4	12.44	85.18	1.90	−14.9	−101	18.2	10.77	−0.2	0.3	1.5	−12.7	−90	11.5
2017	Mar	174.5	16.40	89.94	1.96	−12.4	−89	9.4	23.08	−1.4	−5	6.4	−11.7	−84	9.9
2017	Apr	118.3	20.25	90.90	1.43	−13.1	−93	11.5	25.11	−2.5	−11	9.1	−12.3	−85	13.5
2017	May	421.6	23.89	88.45	1.72	−13.4	−94	12.6	28.15	−4.4	−23	12.5	−14.5	−101	15.1
2017	Jun	406.2	27.82	89.17	1.81	−14.2	−103	11.2	37.17	−6.9	−47	8.6	−15.7	−112	13.6
2017	Jul	312.2	27.85	86.68	1.45	−13.8	−97	13.3	35.59	−6.0	−40	7.5	−17.4	−123	16.7
2017	Aug	245.2	28.00	87.39	2.05	−18.5	−134	13.8	35.49	−8.3	−55	11.1	−18.8	−133	17.7
2017	Sept	270.2	26.84	91.10	1.30	−16.0	−118	10.2	36.41	−6.5	−42	10.1	−16.1	−113	15.2
2017	Oct	44.3	22.59	77.45	2.31	−14.2	−102	11.5	27.40	−4.8	−30	8.4	−15.1	−110	11.4
2017	Nov	37.1	17.76	83.60	2.62	−15.7	−114	11.3	24.50	−3.5	−24	4.2	−15.1	−110	11.3
2017	Dec	1.4	12.35	70.39	2.75	−16.6	−93	34.9	13.83
2018	Jan	132	11.20	89.13	2.54	−15.0	−107	12.8	15.26	−5.4	−29	14.3	−15.4	−111	12.3
2018	Feb	12.6	11.67	77.32	2.11	−12.9	−94	9.2	16.03	−1.8	−3	12.0	−12.5	−91	9.4
2018	Mar	56.4	19.36	80.32	2.31	−12.6	−91	9.8	21.30	−0.7	−2	3.2	−12.4	−86	13.7
2018	Apr	137.2	21.76	82.76	2.07	−13.3	−95	11.3	18.94	−0.4	−1	2.8	−11.3	−79	11.6
2018	May	98.9	27.62	80.70	2.15	−13.3	−97	9.3	29.36	−1.0	−5	3.3	−13.0	−89	14.7
2018	Jun	489.1	27.44	85.27	1.98	−15.9	−114	13.4	34.32	−8.1	−53	11.4	−18.2	−127	19.1
2018	Jul	270.9	28.48	84.06	2.05					−3.1	−23	1.3	−13.8	−104	6.5
2018	Aug	218	27.90	87.23	1.80					−7.4	−55	3.8	−16.1	−122	6.1
2018	Sept	155.3	27.04	83.70	2.20	−17.9	−137	6.2	21.6	−8.6	−69	−0.6	−17.0	−130	5.5
2018	Oct	82.3	22.36	78.61	2.28	−17.1	−128	9.4	20.06	−2.8	−14	8.3	−14.3	−99	15.3
2018	Dec	18.7	15.7	80.8	2.9	−16.5	−106	26.0	21.48	−1.8	−10	5.1	−13.1	−95	9.6

). Relative humidity varied from 36% to 100%, averaging 85%. Relative humidity exceeded 77% for more than 270 days annually. Monthly average relative humidity was observed to be higher than 80% from March to September, with the lowest value recorded in February. The total precipitation during the study period was 6213.4 mm, 75% of which occurred from April to September. Consequently, the wet season is defined as April to September, and the dry season as November to February. March and October were considered transition periods for seasonal change. Wind speed varied from 0.04 m/s to 8.70 m/s, with an average of 2.00 m/s. Lower values were observed during the wet season, while higher values were recorded in the dry season, with a maximum fluctuation of 5.82 m/s.

3.2. The Isotopic Composition of Water Vapor and Precipitation

The daily δ¹⁸O_v values in the wet season ranged from −24.6‰ to −9.9‰, with an amplitude of 14.6‰. In contrast, during the dry season, values ranged from −19.9‰ to −11.1‰, with an amplitude of 8.9‰. The variability of δ_v during the wet season was greater than that observed in other periods. This observation is consistent with findings from Beijing [10], New Haven [28], and Japan [29]. Similar variations were observed for δD compared to δ¹⁸O during the study period.

Monthly δ¹⁸O_v values exhibited an increase prior to the onset of the wet season (Table 1). A continuous decrease was observed from −12.1‰ in April to −18.6‰ (−13.1‰) in September 2016. Subsequently, an increase was noted from October to December, followed by another decrease during the early summer monsoon period. Similar systematic seasonal variations in δ_v were also observed for 2017 and 2018. The values of δ_v also exhibit enrichment during the dry season in high-latitude regions, such as the Antarctic Plateau [30] and the Tibetan Plateau [31].

Monthly δ¹⁸O_r values decreased from −1.3‰ in April to −5.1‰ in September, subsequently increasing from October 2016. Consistent with previous studies, depletion of light isotopes in precipitation during summer and enrichment during winter have been observed in Guangzhou [32,33].

Both δ_v and δ_r exhibited systematic seasonal changes, which were influenced by monsoonal patterns. The isotopic characteristics of water vary considerably across different regions. Seasonal variations in δ_v have been observed across a wide range of climatic zones, from tropical savanna climates (e.g., Darwin, Australia) [16] to humid continental climates (e.g., Kourovka, Russia) [34].

The δ¹⁸O_e values ranged from −23.5‰ to −7.5‰, with an average of −15.0‰. The variability during the wet season (−14.6‰) was approximately 1.5 times greater than that observed during the dry season (−10.8‰).

3.3. Characteristics of LMWL and LMVL

The Local Meteoric Water Line (LMWL) was estimated based on the daily values of δ_r (Figure 2) and is expressed by the equation δD = (7.9 ± 0.1) δ¹⁸O + (9.4 ± 0.8) (N = 239, R² = 0.95). The slope and intercept of this LMWL were consistent with the isotopic characteristics typically observed in low-latitude regions characterized by high temperatures, high relative humidity, and low altitudes. The Local Meteoric Vapor Line (LMVL), derived from daily isotope ratios of water vapor, was δD = (6.5 ± 0.1) δ¹⁸O − (9.5 ± 16.5) (N = 386, R² = 0.90). The LMVL exhibited a smaller slope and a negative intercept when compared to the LMWL. The negative intercept of the LMVL diverges from the simulations of several general circulation models, including ECHAM, GISS, LMDZ, and isoGSM [14,15,35,36]. Presently, isotopic data for water vapor provided by GNIP and other published works consistently show negative intercepts for LMVLs, such as −1.43 for Vienna, −10.81 for Rehovot, and −3.13 for Beijing. The Equilibrium Local Meteoric Vapor Line (ELMVL) was determined to be δD = (7.1 ± 0.1) δ¹⁸O − (0.9 ± 1.6) (N = 239, R² = 0.95). The intercept of the ELMVL was also found to be negative.

4. Discussion

4.1. Isotope Effect of Precipitation and Water Vapor

A negative correlation was observed between the daily value of δ¹⁸O_r and precipitation amount (P) during the wet season (r = −0.21, p < 0.01), whereas a positive correlation (r = 0.11) occurred in the dry season (

Table 2. Pearson correlation analysis for isotope ratios of water vapor and precipitation.

Meteorological		Precipitation			Vapor
Elements		δ18Or	δDr	dr	δ18Ov	δDv	dv
P (mm)	WS	−0.21 **	−0.18 *	0.14	−0.16 *	−0.15 *	0.09
	DS	−0.11	−0.05	0.13	0.02	0.01	−0.06
	YR	−0.22 **	−0.20 **	0.11	−0.10 *	−0.13 *	−0.02
T (°C)	WS	−0.39 **	−0.44 **	−0.22 **	−0.08	−0.15 *	−0.18 **
	DS	−0.19	−0.34 *	−0.42 **	0.33 **	−0.19	−0.35 **
	YR	−0.42 **	−0.50 **	−0.37 **	−0.01	−0.11 *	−0.28 **
RH (%)	WS	−0.03	−0.02	0.14	0.12	0.13	−0.04
	DS	0.03	0.03	0.02	0.11	0.11	−0.1
	YR	−0.01	0.01	0.11	0.18 **	0.09	−0.30 **
V (m/s)	WS	−0.14	−0.16 *	−0.1	−0.18 **	−0.20 **	0.04
	DS	−0.11	−0.1	0.01	−0.11	−0.11	0.03
	YR	−0.05	−0.06	−0.02	−0.27 **	−0.25 **	0.16 **

Note: WS is short for wet season, DS is short for dry season, YR is short for year; * significant at p < 0.05, ** means at p < 0.01.

). On an annual scale, a negative correlation was observed between δ¹⁸O_r and P (r = −0.22, p < 0.01). These results suggested that the daily values of δ_r in Guangzhou do not exhibit a pronounced amount effect. Interestingly, an “inverse amount effect” during the wet season has been reported for certain subtropical and tropical regions, such as observations from Guam and New Delhi [37]. Conventional interpretations of the amount effect have often failed to adequately explain isotopic variations in these climates. In such regions, precipitation is predominantly derived from deep convection, where vertical movement dominates over horizontal transport. Consequently, the amount effect can be modified by convection or large-scale advection processes [35,38].

A significant negative correlation between δ¹⁸O_r and temperature (T) was consistently detected throughout the year, as well as in both the wet and dry seasons. This relationship is in opposition to the anticipated temperature effect. This finding is contrary to the expected temperature effect. This pattern is similar to that observed for summer precipitation isotope ratios in middle- latitude and low-latitude coastal or monsoon regions in eastern China [39]. Spatial variations in the temperature effect may be attributed to eddy diffusion and large-scale advection influencing water vapor transport [38].

The intensity of kinetic fractionation during evaporation is primarily controlled by relative humidity (RH) and wind speed (V). Low relative humidity provides the driving force for evaporation, whereas high wind speed enhances and sustains this fractionation by disrupting the saturated vapor layer above the liquid surface [40]. These factors jointly determine the isotopic composition of source-region water vapor and ultimately influence the isotopic ratios observed in precipitation. In New Haven, RH, rather than T, has been identified as a superior predictor for changes in δ¹⁸O_r on short timescales (e.g., less than a few weeks) [28]. But no significant correlation was demonstrated between δ¹⁸O_r and either RH or V in Guangzhou.

During the wet season, the correlation between the daily values of δ¹⁸O_v and P was relatively weak (r = −0.16, p < 0.05). On the annual scale, the negative correlation between δ¹⁸O_v and P was even weaker (r = −0.10, p < 0.05). Similarly to the δ_r, no clear amount effect was detected for δ_v in Guangzhou. However, when the monthly weighted average values of δ_v were utilized, the correlation coefficient between them increased to −0.23 (p < 0.05). Previous studies have also reported amount effects in δ_v, although these can be modified by convection [41,42]. This suggests that the amount effect between the δ_v and δ_r needs to be established at a scale longer than the monthly scale. Long-term observations of δ_v are required to establish the relationship between δ_v and atmospheric processes such as condensation, convection, and atmospheric circulation.

At both the annual scale and during the wet season, the daily values of δ¹⁸O_v showed weak correlations with T (r = −0.08 and −0.01, respectively). The δ¹⁸O_v values in the dry season were positively correlated with T (r = 0.33, p < 0.01), indicating that the temperature effect on the values of δ_v is active during the dry season in Guangzhou. It is recognized that the relationship between the δ_v and T exhibits both temporal and spatial variability [37,43]. One explanation is that temperature effect can be altered by mixing processes along trajectories and by isotope exchange with the lower surface [44]. During the wet season, the extensive scouring by large amounts of precipitation, coupled with the mixing of surface evaporation under high T and high RH conditions, appears to obscure the temperature effect. An alternative explanation is that the values of δ_v is likely related to the last saturation temperature, rather than the local temperature [45,46].

A weak but significant positive correlation was observed between δ¹⁸O_v and RH (r = 0.18, p < 0.01) on an annual scale. This relationship reflects the influence of local weather conditions in dry season [40,47]. A correlation between RH and δ_v has been observed in both subtropical and high-latitude regions. For instance, a significant correlation exists between δ_v and RH in Greenland [48]. This suggests that the seasonal variations in δ_v are related to local weather conditions. In East Asia, the relationship between δ_v and RH is attributed to vertical moisture mixing, which can induce changes in water vapor mixing concentration, consequently leading to variations in near-surface δ_v [7]. Under conditions of low RH, the depletion of heavy isotopes from surface evapotranspired water vapor would decrease the overall δ_v, leading to a significant positive correlation between d_v and RH (r = −0.30, p < 0.01).

The correlation coefficient between V and δ¹⁸O_v was −0.27 (p < 0.01) throughout the year, but it weakened to −0.18 (p < 0.01) in the wet season. For coastal regions, V may represent one of the most critical factors influencing δ_v. High V can enhance turbulent mixing within the atmospheric boundary layer, promoting cyclical variations in δ_v [49]. Strong convective are also a significant factor that limits variations in δ_v in coastal areas. Increased subsidence from high-pressure systems facilitates the transport of dry air from the upper atmosphere to the surface, which is typically associated with calm or low-wind conditions [30]. Changes in water vapor mixing concentration occur under these conditions, leading to variations in δ_v.

4.2. Relationship and Deviation Between Isotope Ratio of Water Vapor and Precipitation

4.2.1. Correlation Controlled by Equilibrium Fractionation

Both equilibrium and kinetic fractionation play critical roles in the variability of stable isotopes in water [1,2,3]. In this study, three complementary approaches were employed to assess whether equilibrium fractionation prevailed during precipitation: (1) the isotopic difference between precipitation and water vapor; (2) the intercept difference between the LMWL and LMVL; and (3) the deviation of observed vapor isotopic composition from its theoretical equilibrium value.

First, the values of δ_v were observed to be substantially lower than those of δ_r. On the annual scale, the deviations between δ_v and δ_r (δ¹⁸O = −9.6‰, δD = −75‰) were found to be very close to the liquid–vapor equilibrium fractionation values under 25 °C conditions (δ¹⁸O = −9.3‰, δD = −74‰). This indicates that the isotopic fractionation signals recorded locally are consistent with those predicted by the equilibrium fractionation process. Furthermore, it would be expected that if kinetic fractionation were a dominant process, considering the mass difference between hydrogen and oxygen isotopes, the deviation between δ_v and δ_e would be expected to exceed that between δ_v and δ_r. However, measurements conducted in Guangzhou during 2016~2018 demonstrated the opposite, as the mean deviations between δ_v and δ_e were only δ¹⁸O = −0.22‰ and δD = −2‰.

Second, the slopes of LMWL and LMVL were found to be 7.90 and 6.45, with intercepts of 9.35 and −9.51, respectively. A key expectation under conditions of ideal equilibrium fractionation is that the intercept of the LMVL would be predicted to be consistent with the intercept of the LMWL, assuming complete vapor saturation accompanies precipitation. In this study, the absolute difference in the intercepts was found to be only 0.06, and the slopes were similar. This minimal deviation indicates that non-equilibrium fractionation during the precipitation process is not the dominant controlling factor, and the system is operating close to ideal equilibrium fractionation. This suggests that the kinetic fractionation experienced by precipitation during its fall, particularly re-evaporation, is a very weak process.

Third, the scatter plot of the δ_v and the δ_e indicates that most samples are distributed near the 1:1 line. This distribution suggests that the local isotopic composition of atmospheric water vapor is in good agreement with the values predicted by equilibrium fractionation.

Deviations from the 1:1 line were found to be primarily associated with precipitation events of less than 5 mm or greater than 50 mm. This observation is consistent with previous research, which has shown larger deviations between δ_v and δ_e occurring during cold months without P and during months with low P [8]. Additional research has further demonstrated that the values of δ_v are sensitive not only to sub-cloud processes during periods of limited precipitation but also to heavy rainfall events [12]. The greater deviations that occur during low-precipitation and high-precipitation events suggest that the values of δ_v near-surface cannot be accurately predicted using the values of δ_e under these conditions. To investigate the potential relationship, an analysis was performed between the values of δ_v and δ_e and P to determine if the deviation would increase or decrease with varying P. However, it was determined that P is not the primary cause of these observed deviations (Figure 3). Instead, the differences between δ_v and δ_e are likely attributable to additional processes, such as shallow convection or turbulence, reversible adiabatic processes, convective entrainment, re-evaporation of raindrops, or post-condensation isotopic exchange, all of which influence local variations in δ_v [7].

4.2.2. Mixing of Ground-Level Water Vapor into the Cloud Base

The observed difference in the intercept (8.6) between the LMVL and the ELMVL, while maintaining similar slope values, can potentially be attributed to the vertical profile of the isotope ratio in water vapor. A water vapor concentration gradient exists between near-surface water vapor and the free atmospheric layer, with varying temperatures at different altitudes. Changes in temperature influence the saturated vapor pressure, leading to the development of a vertical profile of vapor isotopes [42].

Previous studies have verified the altitude effect on the isotope ratio in precipitation [2,50]. Obtaining high-resolution data on the altitude effect of the isotope ratio in water vapor typically necessitates the use of aircraft and remote sensing instruments. Given the limitations of the available sampling equipment, it was assumed that the isotope ratio of water vapor exhibits an altitude effect similar to that of precipitation. The measured isotope ratio of near-surface water vapor was utilized as the basis for a simplified estimation of the water vapor isotope ratio at different heights, which can be calculated using Equation (3). Figure 4a illustrates the differences in δ¹⁸O and δD between upper-tropospheric water vapor and near-surface water vapor, expressed as Δδ¹⁸O and ΔδD, respectively. For every 500 m increase in height/altitude, the change in δ¹⁸O_v was approximately −0.05‰, while the change in δD_v was about −4‰. As height increases, the absolute value of ΔδD progressively increases, resulting in a corresponding increase in the absolute value of the LMVL intercept. In other words, the discrepancy in the intercept between the LMVL and ELMVL was caused by the differential variation of δ¹⁸O and δD with height.

The LMVL at different heights is presented in Figure 4b. As height increases, the difference between the intercept of the LMVL and ELMVL becomes more pronounced. The intercept decreased from −9.5 at an altitude of 500 m to −16.5 at 3000 m, which is consistent with other reported results [3,51]. A steep vertical gradient of δ_v is known to exist, primarily due to the background thermal structure and its influence on condensation history [42]. Consequently, the distinct intercepts of the LMVL and ELMVL are considered to result from differences in the vertical profiles of δ_v.

Furthermore, it can be observed that the ELMVL was most closely aligned with the LMVL corresponding to altitudes of 750–1000 m. When the equilibrium isotope ratio of water vapor matches the isotope ratio of water vapor below the cloud base, it can be interpreted that the deviation between δ_e and δ_v is a result of the mixing of water vapor from the surface to the cloud base. This suggests that local precipitation typically occurs at an altitude of 750–1000 m. In fact, the average cloud base height during summer in South China ranges approximately from 400 to 1000 m, with convective precipitation being predominant in spring and summer. In contrast, multi-layer cloud precipitation is dominant in winter, with average cloud base heights exceeding 1000 m [52].

4.3. Identification of Local Water Vapor Sources

The relationship between water vapor concentration (q) and the isotope ratio of water vapor (δD) is considered a fundamental aspect of the two-end-member mixing model in water vapor isotope research [53]. The combination of the Rayleigh curve, the mixing curve, and the δD − q relationship can be employed to identify the sources of atmospheric water vapor. The δD − q relation (Figure 5a) was derived from Equation (4) and water vapor measurements obtained from space-borne platforms [5,9]. Both the Rayleigh curve and the mixing curves were established as functions of water vapor concentration.

Water vapor concentration (q) within the 500–825 hPa layer typically ranged between 0 and 10 mmol/mol [53,54]. The total vertical structure of δD in tropical regions has been characterized, showing a minimum value of −675‰ near 15 km and nearly constant values maintained above the stratosphere [53,54]. Unlike the satellite-derived results, q in Guangzhou were considerably higher, often exceeding 20 mmol/mol and, in some cases, even 30 mmol/mol, with δD values predominantly around −100‰. In Guangzhou, the values of δ_v during the wet season has been observed to closely follow the Rayleigh curve, owing to its proximity to oceanic moisture sources. By contrast, in the dry season, when water vapor is primarily derived from a mixture of both oceanic and continental sources, the isotopic signatures tend to deviate from the Rayleigh curve. In a mixed state with the same q, the values of δ_v resulting from mixing is expected to be significantly larger than that predicted by Rayleigh fractionation/distillation [53].

Over 38% of the values of δ_v exhibited characteristics consistent with mixing with oceanic sources, falling near the mixing curve for such sources. Water vapor characterized by high temperature and high humidity plotted on the oceanic source mixing curve, indicating either rapid movement after evaporation from the ocean or significant turbulent mixing [7]. Most subtropical areas experience reversible adiabatic exchange, and water vapor within the oceanic boundary layer is dominant before mixing with local water vapor [47,53,54].

Points falling between the oceanic source mixing curve and the Rayleigh curve, which accounted for 8% of the observations, could be attributed to post-condensation processes such as re-evaporation or mixing with surrounding air. The subtropical region near the ocean is characterized as an active convection area. The transport and mixing processes of vertical air masses can change within 1–3 days. Such short-term variations may be caused by water vapor interference with local or other regional atmospheric masses [47,53]. Condensed precipitation is a process during which water vapor undergoes dramatic changes. Evaporation occurring below the cloud base would lead to changes in q. Furthermore, strong convection in subtropical and tropical regions often results in mixing with surrounding air [35].

For values falling below the Rayleigh curve, predominantly observed during June-July-August (JJA) and September-October-November (SON), the amount effect associated with remoisture processes occurring below the cloud base is understood to produce a δ_v value lower than that expected by Rayleigh fractionation [53,54]. Satellite observations have indicated that an increase in precipitation rate can cause a portion of the values of δ_v to fall below the Rayleigh curve [7], a finding consistent with our observations. As Guangzhou is characterized by an East Asian monsoon climate with two flood seasons (MAM and JJA), typhoons frequently occur during the JJA, contributing to 36% of the total precipitation. A short period of heavy precipitation is known to cause severe depletion of heavy isotopes in water vapor, resulting in values falling below the Rayleigh curve.

During the dry season, 7% of the δ_v values were located in the region indicative of mixing between land and ocean water vapor sources. The distribution of air trajectories simulated by HYSPLIT was found to be consistent with the δD−q relationship. Although the contribution from winter continental vapor trajectories increased to 39%, this was insufficient to diminish the predominance of the oceanic vapor source. Furthermore, strong winter winds are known to facilitate post-condensation processes, such as the re-evaporation of raindrops and their mixing with ambient air. These processes enhance isotopic kinetic fractionation, which contributes to greater deviations between the observed δ_v and the values predicted by the Rayleigh fractionation model.

The points situated between the oceanic source mixing curve and the Rayleigh curve originated from months marking the transition between the rainy and dry seasons. During these transitions, air trajectories from the ocean sharply declined, while those from westerlies and the northern mainland increased, and vice versa. It is important to note that the local δD−q relationship did not feature points indicative of mixing with land sources. This absence is attributed to the fact that only values of δ_v collected during precipitation events were selected, a period when the contribution from land water vapor sources was minimal.

5. Conclusions

Measurements were presented for hydrogen and oxygen stable isotopes in atmospheric water vapor (δ_v) near-surface and precipitation (δ_r). Combined with simultaneous meteorological data, precipitation processes and isotopic fractionation during phase transitions were explored. The variations in the values of δ_v are consistent with those observed in the values of δ_r, exhibiting systematic seasonal pattern characterized by heavy isotope depletion during the wet season and heavy isotope enrichment during the dry season.

Atmospheric circulation serves as the underlying control for the temporal variations in the values of δ_v and δ_r. The variability in the values of δ_v during the wet season was found to be smaller than that in the dry season, whereas the opposite trend was observed for the variability in values of δ_r. The values of δ_v are significantly influenced by meteorological factors, with the dominant factors varying across seasons. Among these, wind speed was found to have the most significant impact on the values of δ¹⁸O_v, whereas relative humidity was identified as a secondary factor influencing the variations of δ¹⁸O_v in coastal regions. Subsidence of water vapor within the boundary layer under conditions of low humidity and low wind speed can lead to the depletion of light isotopes in water vapor.

A monthly amount effect was observed for δ_v, but no such effect was evident for δ_r. The observed amount effect on the values of δ_v is more likely attributable to water vapor mixing under conditions of strong convection, high humidity, and high wind speed. Long-term observations of δ_v are required to establish the relationship between δ_v and atmospheric processes.

On an annual scale, the deviations between δ_v and δ_r (−9.7‰ for δ¹⁸O and −76‰ for δD) was found to be very close to the liquid–vapor equilibrium fractionation values under 25 °C conditions (δ¹⁸O = −9.3‰, δD = −74‰). This finding indicates a high degree of consistency between the δ_v and the predictions of equilibrium fractionation in Guangzhou.

The slopes of the Local Meteoric Vapor Line (LMVL) and Local Meteoric Water Line (LMWL) are very similar, but the negative intercept of the LMVL represents a distinctive feature of δ_v in this region. These results present a contrast to those from isoGSM models, which typically predict a positive LMVL intercept.

The Equilibrium Local Meteoric Vapor Line (ELMVL) exhibited a similar slope to the LMVL but with an intercept difference of 8.6. The vertical distribution of δ_v is governed by the gradient of water vapor concentration between the near-surface and the free atmosphere, in combination with temperature variations at different altitudes. These factors collectively result in the differences between δ_v and δ_e, as well as the difference between the intercepts of the ELMVL and the LMVL.

Rayleigh fractionation typically considers changes only in the horizontal direction, without accounting for vertical variations. Utilizing the δD − q relationship, it was determined that the majority of water vapor (75%) originates from mixing with oceanic sources during the two local flood seasons. This finding is consistent with HYSPLIT simulations, which indicated that the sum of water vapor trajectories originating from the Indian Ocean, the South China Sea, and the Pacific Ocean constituted 71%. Points situated between the oceanic source mixing curve and the Rayleigh curve are indicative of the interaction between land and ocean water vapor, representing the outcome of water cycle processes during post-condensation. Heavy precipitation and rapid remoisture processes over short periods promote the depletion of water vapor isotope ratios, causing them to fall below the Rayleigh curve.

These results address existing gaps in the understanding of water vapor isotope ratios in East Asia. Furthermore, they contribute to diagnosing deficiencies in isoGSM models concerning stable water isotopes and facilitate the comparison with atmospheric vapor isotopic observations obtained from satellites.