Sri Lanka (located between approximately 6 to 10° N and 79 to 82° E) is the southernmost country on the Indian subcontinent and a key region for identifying the moisture source of the South Asian summer monsoon (Lekshmy et al., 2015). Featuring a tropical climate, Sri Lanka experiences four distinct monsoon seasons annually: the northeast monsoon from December to February, the first inter-monsoon from March to April, the southwest monsoon from May to September, and the second inter-monsoon from October to November (Malmgren et al., 2003; Jayasena et al., 2008). For the analyses, we combined the first and second inter-monsoon periods into a single non-monsoon period. Most of the precipitation in Sri Lanka comes from the southwest and northeast monsoon systems, accounting for over 78 % of the total annual precipitation (Fig. 1c). Precipitation formation in Sri Lanka primarily relies on organized convection associated with the Intertropical Convergence Zone (ITCZ) and low-pressure systems (Gadgil, 2003), while the associated moisture primarily originates in the Indian Ocean and BoB (Bandara et al., 2022). The southwest monsoon transports moisture from the Indian Ocean to southwestern Sri Lanka (Fig. 1a), where it leads to increased rainfall (Rahul et al., 2016b). In contrast, the northeast monsoon carries water vapor from the BoB to northern and northeastern Sri Lanka (Fig. 1b) (Dhar and Rakhecha, 1983; Chang et al., 2006).

An automated weather station (AWS) was installed at the University of Ruhuna, Matara (located at 5.94° N, 80.57° E), on the southern coast of Sri Lanka (Fig. 1d). It collected real-time meteorological observations, including air temperature, precipitation, relative humidity, air pressure, wind speed, and wind direction from 1 March 2020 to 28 February 2021.

In this study, we used hourly data of 2 m air temperature, 2 m dew temperature, air pressure, precipitation, SST, atmospheric boundary layer height (BLH), wind speed, wind direction, and outgoing longwave radiation (OLR), obtained from ERA5 for the years 2000 to 2021, with a spatial resolution of 0.25° × 0.25° and a temporal resolution of 1 h. Meteorological data are compared with water vapor isotopic data measured during the same period. Based on the European Centre for Medium-Range Weather Forecasts (ECMWF, https://cds.climate.copernicus.eu/eu/, last access: 4 May 2024) reanalysis dataset (ERA5), the annual average precipitation and air temperature for the period from 2000 to 2020 are 2085 mm and 27.6 °C, respectively (Fig. 1c) (Hersbach et al., 2020). Studies have shown that ERA5 data provide good representations of the Matara equatorial climate and can be used in lieu of missing observational data (Fig. S3 in the Supplement) (Bandara et al., 2022). Due to weather conditions and instrument trouble, specific humidity measured by the isotopic measurement instrument and computed by the AWS are missing from March to April 2020. Therefore, we chose to present both variables obtained from ERA5 as they complement each other, providing a clearer picture of humidity changes at Matara station.

For the atmosphere above open sea regions, RH_SST is obtained using the following (Bonne et al., 2019): 1RHSST=RH2mair×qsatT2mairqsatSST, where RH_2 m air is the relative humidity 2 m above the ocean surface, qsatT2mair is the specific humidity at a saturated condition for a given 2 m air temperature, and qsatSST is the specific humidity calculated for seawater with a salinity of 35 practical salinity units (PSU) (Curry and Webster, 1999).

The formulas to calculate air saturation specific humidity qsatT2mair and sea-surface-saturation specific humidity qsatSST with a salinity of 35 PSU are the following: 2qsatT2mair=0.622×EP,3qsatSST=0.98×qs, where qs represents specific humidity with a salinity of 35 PSU and is calculated in the same way as qsatT2mair. E is the saturated water vapor pressure, obtained from the improved Goff–Gratch formula (Goff and Gratch, 1946). P is atmospheric pressure. We take the sea surface pressure as the atmospheric pressure to participate in the above calculation (Eq. 2).

Near-surface atmospheric water vapor isotope measurements at Matara were collected using a sampling frequency of 1Hz with the instrument located approximately 5m from the AWS. We used a water vapor isotope analyzer (Los Gatos Research (LGR) Inc.) in conjunction with an LGR water vapor isotope standard source (WVISS) model. The LGR instrument employs a mirrored sampling chamber in which the laser traverses the sample volume thousands of times, effectively amplifying the water vapor absorption signal which facilitates the detection of low concentrations of D and ¹⁸O (Liu et al., 2015). Compared to traditional methods, this spectroscopic technique offers three advantages: (i) it is compact and portable, enabling real-time field monitoring; (ii) it can simultaneously measure δ18O and δD; and (iii) it has lower measurement costs and requires less operator expertise.

The instrumental setup was situated approximately 100 m from the sea shore (5.94° N, 80.57° E; 10 m) and consisted of four primary components: (1) a sampling inlet, positioned approximately 5 m above the ground (Fig. 1d), equipped with a stainless-steel mesh to prevent interference from insects and facing downward to avoid direct impacts from rain; (2) a 1/4 in. outer diameter stainless-steel sampling tube, insulated with heating tape and a 2 cm thick pipe for thermal insulation; (3) the calibration unit to generate a constant flow of water vapor with known isotopic composition and at different humidity levels; and (4) a water vapor isotope analyzer, delivering measurement precisions for δ18O and δD of 0.25 ‰ and 0.5 ‰, respectively (a concentration of 2500 ppmv). This setup has been designed to minimize external influences and maintain the integrity of the sampled water vapor.

The spatial proximity between the water vapor analyzer and AWS ensures a high level of synchronicity between the isotope and meteorological measurements. We define wind directions from 60 to 330° N as oceanic, while those from 330 to 60° N are defined as terrestrial winds (Fig. 1a, b).

Atmospheric water vapor stable isotopes are expressed using the δ notation (in per mil, ‰), using the following equations: 4R18O=18O16O,5RD=D1H,6δsample=RsampleRVSMOW-1×1000‰. Here, δsample represents either δ18O or δD (¹⁸O or D isotope ratio) relative to Vienna Standard Mean Ocean Water (VSMOW). Rsample and RVSMOW are the ¹⁸O or D and VSMOW isotope ratios, respectively.

In this study, we adhere to the calibration protocol proposed by Steen-Larsen et al. (2013). Briefly, the instrument calibration and data processing consist of three major steps: (1) humidity–isotope response calibration, (2) VSMOW and Standard Light Antarctic Precipitation (VSMOW-SLAP) calibration, and (3) drift correction (see Sect. S1 in the Supplement).

The water vapor concentration can influence the measured water vapor isotopic composition, known as concentration- or humidity-isotope dependency characterization. By adding a constant stream of water vapor with known isotopic composition at different humidity levels, we can establish the humidity-isotope response function (Sturm and Knohl, 2010; Aemisegger et al., 2012). As this function can vary over time, its calibration was repeated monthly, using two standard samples of known isotopic compositions measured at humidity levels ranging from 16 000 to 38 000 ppmv at intervals of 1000 ppmv. Each level was measured for at least 25 min using the LGR WVISS. Our results are referenced to a humidity level of 20 000 ppmv. We compared our measurements to the international VSMOW-SLAP scale, assuming a linear drift between calibration points.

To compensate for instrumental drift, we measured the water vapor from a drift–standard bottle for 25 min after each 12 h performed an ambient air measurement. Furthermore, we tested for instrument drift as part of the routine instrument maintenance, assuming a linear drift between each drift-standard measurement. Laboratory analyses of liquid isotopes have confirmed the stability of its isotopic composition over time.

The Rayleigh distillation model is employed to quantify isotopic variations during phase changes (Dansgaard, 1964), by which the residual air mass becomes drier with a depletion in heavy isotopes following moist adiabatic vertical ascent (Gat, 1996): 7Rr=R0fαvl(T)-1. Here, Rr and R0 represent the isotopic ratio of residual vapor and initial vapor, respectively. αvl(T) denotes the equilibrium fractionation factor, and f is the fraction of residual water vapor.

By integrating the definition of isotope ratios as given in Eq. (6), the Rayleigh distillation model can be expressed in terms of isotopic content as follows: 8δr=δ0+1fαvl(T)-1-1, where δr and δ0 are the isotope ratios relative to VSMOW in residual and initial vapor, respectively.

We employ the mixing model to examine the isotopic characteristics after the mixing of two air masses (Galewsky and Hurley, 2010): 9Rmix=fHDO1+1-f×HDO2fH2O1+1-f×H2O2, 10Rmix=fH218O1+1-f×H218O2fH2O1+1-f×H2O2, where Rmix represents the isotopic ratio of the mixed air mass, while [HDO], [H2O], and [H218O] denote isotopic water vapor volume mixing ratios. f is the mixing fraction. The isotopic ratio and isotopic δ in Eqs. (9) and (10) have been calibrated by VSMOW.

We use water vapor isotopes to characterize the mixing processes in the marine boundary layer (MBL) (Benetti et al., 2018), using the following equation (Craig and Gordon, 1965): 111+δe=1αk×αvl×1+δOC-RHSST×1+δMBL1-RHSST, where αvl represents the equilibrium fractionation factor between vapor and liquid, and αk is the kinetic fractionation factor. δOC denotes the isotopic composition of the ocean surface. We utilize αvl from Wahl et al. (2022) and αk for the smooth regime (αk18O = 1.006 and αkD = 1.0053) (Merlivat and Jouzel, 1979).

To delineate water vapor transport paths and pinpoint moisture sources, we employed the HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model from the US National Oceanic and Atmospheric Administration (NOAA) to compute backward trajectories of air masses arriving at Matara station during the southwest and northeast monsoons. The Global Data Assimilation System (GDAS) with 1° × 1° and 3 h spatial and temporal resolutions provided the background meteorological data from May to September 2020 and December 2020 to February 2021 (ftp://arlftp.arlhq.noaa.gov/archives/gdas1/, last access:19 January 2024). The HYSPLIT model uses GDAS reanalysis data, which contains 37 (vertical) pressure levels and a 1° × 1° horizontal resolution. Atmospheric water vapor primarily resides at altitudes below 2 km (Wallace and Hobbs, 2006). In this study, particles were released four times daily (at 00:00, 06:00, 12:00, and 18:00 UTC) at 20 different locations within a rectangular area extending 0.2° in each direction (north, south, east, and west) from Matara station and at four heights above the ground (50, 500, 1200, and 2000 m). Each trajectory was back-traced for 168 h, recording data at 1 h intervals. The HYSPLIT model outputs latitude, longitude, elevation, pressure, temperature, precipitation, relative humidity, and specific humidity. Backward trajectory clustering analysis was conducted, using the corresponding meteorological data. We averaged the trajectories of four times per day to obtain a daily mean trajectory, combined with water vapor stable isotope values on precipitation days. These daily mean trajectories were clustered by moisture source using K-means clustering. By analyzing the variations in latitude, elevation, and specific humidity along the trajectories, the influence of different moisture sources on local vapor content and isotopic composition was evaluated.

These analyses yielded concentration-weighted trajectory (CWT) fields (resolution of 0.5° × 0.5°) (Hsu et al., 2003) using the in situ daily average δ18O and d-excess, which in turn facilitated the identification of potential moisture sources and an assessment of the potential influence of recirculation on d-excess in water vapor (Salamalikis et al., 2015; Bedaso and Wu, 2020; Xu et al., 2022). CWT (Ci,j) was calculated as 12Ci,j=∑k=1KCkτi,j,k∑k=1Kτi,j,k, where the set (i,j) denotes grid coordinates, k the trajectory index, K the total number of trajectories analyzed, Ck the concentration (here δ18O and d-excess) at the end of the trajectory k, and τi,j,k the residence time of trajectory k in grid cell (i,j). We substituted the residence time by the number of trajectory endpoints in each grid cell (i,j).

During a year-long observational period (1 March 2020 to 28 February 2021), the seasonal and synoptic variations observed in water vapor isotopes (δ18O, δD, and d-excess) and key meteorological parameters (temperature, relative humidity, atmospheric pressure, specific humidity, and SST) are explored (Fig. 2), aiming to provide insights into the interactions between atmospheric and oceanic conditions at Matara. The tropical setting at Matara leads to relatively modest temperature fluctuations, with the 12-month average at 27.6 °C, dipping to 22.3 °C and reaching 33.5 °C at extremes. The amplitude of temperature remains around 10 °C between monsoon and non-monsoon periods. Specific humidity is ranging from 16 to 21 g kg⁻¹ and shows marked oscillations during the southwest and northeast monsoons, with amplitudes of approximately 1.3 and 2.3 g kg⁻¹, respectively. Relative humidity also shows clear seasonal pattern, peaking at 95 % in May (southwest monsoon) and falling to 49.2 % in January (northeast monsoon). Monthly trends (Fig. S4 and Table 1) reveal a steady decrease in both air temperature and specific humidity from May to September, culminating in their lowest values (26.9 °C and 18.5 g kg⁻¹, respectively). During the southwest monsoon, temperatures and specific humidity in May (monthly averages of 28.4 ± 1.4 °C and 21.0 ± 1.1 g kg⁻¹) reflect the influx of warm, moist air from the Indian Ocean. In contrast, February stands out as the coldest and driest period (27.4 ± 2.6 °C and 17.1 ± 1.3 g kg⁻¹), corresponding to the northeast monsoon (Fig. S4). Among them, daily average SSTs consistently exceed the daily average 2 m air temperatures recorded by the AWS, indicating the ocean's significant role in modulating near-surface atmospheric conditions (Fig. 2).

Yearly averages for water vapor isotopic values are -11.6 ‰ for δ18O, -79.5 ‰ for δD, and 13.3 ‰ for d-excess, respectively. Isotopic composition ranges from -23.9 ‰ to -7.5 ‰ for δ18O, -173.2 ‰ to -53.4 ‰ for δD, and -1.2 ‰ to 28.1 ‰ for d-excess (Table 1). Monthly averages of water vapor isotopes (δ18O and d-excess) exhibit stability from March to October, followed by sudden decreases. δ18O and δD show distinct seasonal variations, with higher values during the southwest monsoon and lower values during the northeast monsoon (Table 1). Therefore, the subsequent analysis will concentrate on the variations in δ18O. During the southwest monsoon, the northeast monsoon, and the non-monsoon periods, the average values of δ18O are -11.1 ‰, -12.2 ‰, and -11.9 ‰, respectively. Extreme values of δ18O are observed during the northeast monsoon, with a maximum of -7.5 ‰ and a minimum of -23.9 ‰. Conversely, d-excess exhibits a reverse pattern to δ18O on both seasonal and monthly scales, characterized by lower values during the southwest monsoon and higher values during the non-monsoon period. Furthermore, during the northeast monsoon, the southwest monsoon, and the non-monsoon periods, the average values of d-excess are 12.4 ‰, 13 ‰, and 14.7 ‰, respectively. The d-excess maximum occurs in November at 28.1 ‰ (monthly average of 15.2 ± 4.3 ‰), while the minimum of -1.2 ‰ was recorded in January (monthly average of 11.3 ± 4.5 ‰). The high values of d-excess are related to moisture recycling. Low specific humidity corresponds to depleted δ18O and elevated d-excess, indicating a strong depletion during the long-distance transport from the source regions to the observation station. Coastal stations such as Bengaluru, Ponmudi, and Wayanad also show similar water vapor isotopic depletion in autumn and winter, reflecting the observations made at Matara (Table 2).

For δ18O, δD, and d-excess, synoptic variations were recorded (Fig. 2). Abrupt changes occurred in late July 2020 and from November 2020 to January 2021, associated with synoptic events. Cumulative precipitation in July 2020 reached 451.8 mm, with a notable rainfall event in late July recording daily rainfall of 93.2 mm. Isotopic δ18O values show a sharp depletion from -10.4 ‰ to -20.4 ‰ within 20 h of isolated rainfall events, lasting for 6 d. Over the 75 d period spanning from the late southwest monsoon to the mid-northeast monsoon, significant fluctuations can be seen in isotopic δ18O between -22 ‰ and -11 ‰. During the southwest monsoon from 12 July to 7 August, δ18O values varied from -20.4 ‰ to -9.2 ‰, and δD values ranged from -143.5 ‰ to -68.6 ‰. This finding is consistent with water vapor isotopic δ18O (-14.1 ‰ to -9.8 ‰) and δD (-97.2 ‰ to 69.1 ‰) values measured from 12 July to 7 August 2012, near the Bay of Bengal, although the local minimum at Matara station is below the minimum in the Bay of Bengal (Midhun et al., 2013).

The atmospheric water vapor line serves as an indicator of the humidity conditions at the vapor source and the fractionation processes along the transport path. The slope reflects the extent of vapor kinetic fractionation, while the intercept indicates the humidity levels at the vapor source. Comparing the local meteoric water line (LMWL) for δ18O and δD with the global meteoric water line (GMWL), we obtain a slope of < 8 during both monsoon periods (Fig. 3a). Seasonal variations are also visible in δ18O and δD distribution patterns. Daily averages of water vapor isotopic δ18O and δD demonstrate a strong correlation (r = 0.96, slope = 7.26) with a lower intercept at 4.68. The LMWL slope and intercept vary significantly between monsoon and non-monsoon seasons, peaking in the northeast monsoon with values of 7.3 and 3.86, and reaching a nadir in the southwest monsoon with 6.93 and 1.18, respectively. This suggests increased humidity over sea surface vapor sources from the northeast to southwest monsoon, attributed to heightened evaporation and reduced dynamic fractionation effects. During the northeast monsoon, LMWL slope and intercept are higher compared to other periods, indicating significant moisture recirculation.

We found a significant negative relationship between d-excess and δ18O, with the rate of change for d-excess with δ18O being -0.68 ‰ ‰⁻¹ (r = -0.55) (Fig. S5a). This rate is below the -0.05 ‰ ‰⁻¹ value recorded at Bengaluru station (Rahul et al., 2016b). Seasonally, the correlation between both variables weakened during the southwest, northeast, and non-monsoon periods, with respective rates of change of -0.94 ‰ ‰⁻¹ (r = -0.49), -0.69 ‰ ‰⁻¹ (r = -0.54), and -0.65 ‰ ‰⁻¹ (r = -0.44). Similar patterns are detected for temperature–d-excess and specific humidity–d-excess correlations, showing gradual increases in the slopes and intercepts of the water vapor line. Moreover, the concentrated distribution of vapor values during the southwest monsoon and the highly scattered distribution during the northeast monsoon are indicative of the corresponding seasonal distributions of the water vapor line.

Plots of q–δ18O (i.e., the theoretical Rayleigh distillation curves, the mixing line, and the MBL mixing curve) were used to assess mixing conditions during the study period (Fig. 3b). During the southwest monsoon, most measurements are clustered between the Rayleigh and mixing curve, indicating isotopic variability dominated by effects of precipitation and moisture mixing process. Limited water vapor measurements are scattered below the Rayleigh fractionation line, implying a discernible impact of raindrop re-evaporation. Similarly, during the non-monsoon period, most measurements lie between the Rayleigh and mixing curves, with only a few located below the Rayleigh line. During the northeast monsoon, δ18O spans from the upper to the lower extreme of the mixing and Rayleigh distillation curves. The measurements substantially deviate from the Rayleigh curve and show a higher depletion than predicted by the Rayleigh model, likely due to the influence of convective processes.

To look for diurnal cycles in isotopic composition and meteorological parameters, we analyzed hourly averages (Fig. 4c–e). All isotopic (δ18O, δD, and d-excess) and meteorological parameters exhibit strong diurnal variations during both monsoon and non-monsoon periods (Fig. 4). Overall, the diurnal variations in local meteorological parameters of solar radiation during the day and the resulting development of a boundary layer led to the increasing of temperature and wind speed between noon and afternoon, accompanied by a decrease in relative humidity due to significant evapotranspiration. At night, surface radiative cooling causes temperatures to drop, resulting in calmer conditions near the surface and gradual air saturation, indicating a relatively stable atmospheric boundary layer. During the southwest monsoon, δ18O, δD, relative humidity, wind speed, specific humidity, and BLH are generally higher than during the northeast and non-monsoon periods, while d-excess is lower. In the early morning, δ18O steadily decreases, reaching a minimum (-11.26 ‰) around sunrise (∼ 09:00 LT local time). Subsequently, it increases throughout the day, peaking (-10.87 ‰) in the afternoon (∼ 15:00 LT), yielding a diurnal fluctuation of merely 0.45 ‰. Increased specific humidity between 10:00 and 14:00 LT coincides with increasing air temperatures and wind speeds and decreasing relative humidity (Fig. 4c–f). BLH peaks between 14:00 and 16:00 LT, slightly later than other meteorological parameters. The same diurnal variations for each parameter were observed during the northeast monsoon, with maximum changes in δ18O and d-excess of 1.1 ‰ and 6.8 ‰, respectively. Specific humidity peaks between 10:00 and 16:00 LT, accompanied by increases in air temperature, wind speed, and BLH. After 16:00 LT, specific humidity decreases alongside isotopic δ values and other meteorological parameters. The d-excess peaks (14.81 ‰) at 09:00 LT and fluctuates until 23:00 LT, contrasting with the period from 04:00 to 09:00 LT (Fig. 4b). The d-excess exhibits a W-shaped variability, reaching similar highs at 09:00 and 21:00 LT. Specific humidity exhibits a diurnal variation that aligns closely with the δ18O pattern, reaching its minimum before sunrise and peaking around midday (10:00–15:00 LT). Between afternoon and evening, specific humidity remains relatively high and stable. The diurnal variation during the southwest and northeast monsoon periods are 1.28 and 2.32 g kg⁻¹, respectively. Similarities with patterns observed at Lena River station in eastern Siberia (Bonne et al., 2020) suggest potential influences from moisture exchange between the atmosphere and ocean surface, particularly during the northeast monsoon.

To be able to explore water vapor isotopic variations in the sea surface boundary layer, we must first understand the processes and factors that affect isotope variations during ocean surface water evaporation. The primary determinant governing water vapor stable isotope shifts across different regions is the regional moisture transport process, characterized by differences in isotopic variations in the moisture source region, variations in meteorological conditions during the evaporation processes, and divergences of the moisture transport pathways (Bonne et al., 2020). Thus, this section aims to identify factors that drive the seasonal variations of near-surface atmospheric water vapor stable isotopes at Matara, including water vapor origin, transmission routes, and sea surface evaporation conditions in the source regions.

To further understand the different seasonal relationships between δ18O, d-excess, and meteorological parameters, we analyzed potential seasonal differences between the main moisture sources using HYSPLIT. Backward trajectories from the southwest and northeast monsoons were spatially clustered and analyzed for changes in air mass heights and specific humidity (Fig. S6), facilitating the identification of air mass origins. The specific humidity values along the path have been gridded, and we define the end points of trajectories as indicative of the moisture sources. Trajectories that reach Matara during the southwest and northeast monsoons have different origins. During the southwest monsoon, wind directions span from 60 to 360°, and the main origin regions are therefore the Arabian Sea (AS) and Indian Ocean (Fig. 5a). Due to the northward movement of the warm South Equatorial Current, these winds gather significant amounts of moisture along the way, bringing heavy rainfall to Matara (Fig. 5a and b). Conversely, during the northeast monsoon, the main wind direction shifts to 0–225 and 330–360°, such that most trajectories originate in northeast India, where specific humidity is lower (overland), and only a short portion of the trajectory passes over the BoB. The long transport distance results in a greater depletion in water vapor isotopes once the air mass arrives at Matara station.

Moisture from all sources shows seasonal variations, with δ18O values higher during the southwest monsoon than during the northeast monsoon. The shift in water vapor source from the AS in May to the southern Indian Ocean in September leads to δ18O-enriched water vapor from August to September. Enhanced convective activity and rainfall during the southwest monsoon result in δ18O depletion, while tropical storms and hurricanes also contribute to δ18O depletion.

To assess how seasonal shifts in moisture sources impact the isotopic composition of water vapor, we analyzed the relationship between specific humidity, isotope variations, and wind direction at Matara station (Fig. S7). During the southwest monsoon, wind directions were predominantly WNW, correlating with peak specific humidity and highest δ18O values and lowest d-excess, suggesting a westerly moisture source. Conversely, air masses from the east exhibited δ18O depletion and higher d-excess. The northeast monsoon brought drier air from the BoB, leading to specific humidity between 14 and 17 g kg⁻¹ and significantly depleted δ18O values. These air masses likely experienced substantial isotopic fractionation during their overland passage.

We also investigated the influence of water vapor flux, evaporation, and precipitation on isotopic variations. The southwest monsoon saw lower evaporation rates compared to precipitation at Matara station, contrasting with the northern Indian Ocean and western BoB where evaporation surpassed precipitation. The northeast monsoon, influenced by moisture from the BoB and South Asia, showed higher evaporation rates and increasing water vapor flux. Overall, the water vapor flux and budget varied markedly between monsoons, with the upstream vapor budget significantly affecting stable isotope changes, particularly δ18O. The southwest monsoon's increased precipitation and moisture transport from the northeast led to enriched δ18O at Matara. In contrast, the northeast monsoon's moisture transport resulted in a washing effect, causing a gradual δ18O depletion due to continuous condensation and fractionation along the transport pathway (Figs. 6 and S8).

The d-excess exhibits similar seasonal variations at Matara station, with lower values during the two monsoon seasons and higher values during the non-monsoon periods (Table 2, Fig. 4). This seasonal variation may stem from changes in relative humidity in the moisture source areas and further modifications during transport.

Ocean evaporation represents the starting point of the phase transformations that occur within the global water cycle. Identifying the isotopic variations and controlling factors of oceanic evaporation is essential for understanding isotopic shifts in the marine boundary layer. Previous coastal observational studies focusing on the marine boundary layer have confirmed a significant association between d-excess and RH_SST (Pfahl and Wernli, 2009; Steen-Larsen et al., 2015). In cases where kinetic fractionation during air mass transport is either absent or minimal, d-excess can serve as an indicator of the moisture source region (Bonne et al., 2014).

The map of the moisture sources (Fig. 5) identified the Indian Ocean and BoB as the main source areas for moisture arriving at Matara station. To gauge the impact of more local influences, we investigated how changes in sea surface meteorological conditions in the sea around Matara station affects near-surface water vapor isotope concentrations (Fig. S9). During the southwest monsoon, RH_SST in region A (located to the south of Matara between 3–7° N and 56–65° E) ranged from 64 % to 86 %, with SST fluctuating between 27.9 and 31.5 °C. During the northeast monsoon, RH_SST in region B (located to the east of Matara between 6–8° N and 82–85°E) ranged from 54 % to 84 %, with SST fluctuating between 28.1 and 29.1 °C. In comparison with the southwest monsoon, RH_SST is slightly lower, accompanied by less-pronounced variability in SST. The rate of change in d-excess under the influence of RH_SST in the BoB (during the northeast monsoon) is -0.34 ‰ %⁻¹. In comparison, the rate of change in d-excess with the RH_SST of the northern Indian Ocean (during the southwest monsoon) is -0.32 ‰ %⁻¹, suggesting that evaporation over the northern Indian Ocean significantly impacts local d-excess. Studies focused on the BoB's sea surface revealed that RH_SST explains only 25 % of the d-excess variation (d-excess = (-0.55 ± 0.14) × RH_SST + (56 ± 12); r = -0.5). The limited variation in relative humidity during the monsoon period led to a low correlation, indicating that monsoon moisture plays a crucial role in the isotopic composition of water vapor in the BoB (Midhun et al., 2013). Conversely, the observed relationship between near-surface water vapor d-excess at Matara and relative humidity in the surrounding oceanic region during the observational period, with correlation coefficients of -0.56 and -0.62 (p < 0.01), respectively (Fig. 7), reveals a marked negative correlation between d-excess and relative humidity in the nearby Indian Ocean and BoB, indicating that water vapor at Matara is predominantly supplied by nearby oceans. Notably, SST amplitude near the Matara station is smaller than the variations in near-surface air temperature (Fig. 2).

In the equatorial tropics, OLR mainly results from convective activity and cloud cover, which can impact the stable isotope composition of precipitation (Ohring et al., 1984; Gao et al., 2013; Guo et al., 2017). Generally, higher OLR values are associated with weaker convective activity. Examining the correlation between stable isotopes of water vapor and OLR helps to understand the impact of convective activities along near-surface trajectories of water vapor stable isotopes at Matara station.

We calculated the spatiotemporal correlation between OLR and precipitation amount using the measured water vapor isotopic compositions at Matara station. Specifically, we calculated the average precipitation amount and average OLR for each grid point by averaging over different numbers of days (n = 1, 2, up to 30) preceding each day. Lower OLR values indicate the presence of deep convective clouds in this region and higher precipitation associated with lower δ values.

Figure 8a shows the strong positive correlation (red regions) between rainfall and δ18O during the southwest monsoon, mainly in the northern BoB and over India. This correlation strengthens and extends over wider areas as n increases from 1 to 5. Additionally, a strong negative correlation is evident in the northern Indian Ocean and southern Arabian Sea, reaching a maximum for n = 2 d. During the northeast monsoon, the spatial correlation distribution differs, with a negative correlation observed over the southern Indian Ocean and BoB (Fig. 8b), reaching a maximum for n = 5 d. Lower OLR values in the Arabian Sea and the northern part of the India Ocean correspond to a decrease in water vapor isotopic δ18O at Matara station (Fig. 8c, d). This pattern indicates that water vapor δ18O during the northeast monsoon period is influenced by convective activities over the southern BoB and Southeast Asian regions. The stronger this convective activity, the more depleted the air reaching Matara in water vapor isotopic δ18O.

To examine the correlation between water vapor isotopic δ18O and local precipitation (Fig. 8e) and OLR (Fig. 8f), we selected a small region of 5° × 5° around Matara and calculated the temporal correlation for all grid points as described above. The results show that the correlation with precipitation is negative during both monsoon seasons as expected. The depletion of low-level water vapor δ18O is related to the transport and deposition of water vapor into the lower atmosphere through convective activity (Kurita, 2013; Midhun et al., 2013; Lekshmy et al., 2014). The air masses are resupplied to the convective system through moisture recycling. This results in a strong correlation between the isotopic composition of water vapor and the convective activity during the previous day (Fig. 8f). Residual water vapor is more depleted in strong convective systems. In our study, the correlation reaches a high value after about 5 d, indicating that the convective activity is sufficiently established to affect the isotopic composition of water vapor. In fact, the correlation (for p < 0.05 and in absolute terms) is high for all n values, with maxima of about 0.37 for n = 2 d during the southwest monsoon and 0.55 for n = 2 to 5 d during the northeast monsoon.

The OLR correlation peaks at smaller timescales (approximately n = 2–5 d; Fig. 8f) than precipitation (n = 3–7 d). We attribute this difference to the effect of cloud distribution on precipitation and OLR. OLR has a stronger response to shallow clouds, while precipitation is more responsive to both deep convective clouds and shallow clouds (Masunaga and Kummerow, 2006; Schumacher, 2006). The OLR minimum occurs when thunderstorm clouds result in more precipitation. Additionally, deep thunderstorm clouds, with short lifetimes and consequently very low OLR (corresponding to highly depleted water vapor isotopic δ), exhibit a short memory effect on the correlation (peak occurs at smaller timescales) (Gambheer and Bhat, 2000).