Seasonal Variations of Modern Precipitation Stable Isotopes over the North Tibetan Plateau and Their Influencing Factors

Abstract

Based on 360 event-based precipitation samples collected at six stations on the North Tibetan Plateau (NTP) in 2019–2020, we analyzed the influence of meteorological parameters, sub-cloud evaporation, moisture sources, and moisture transmission pathways on precipitation and its seasonal variations. The results show that precipitation δ¹⁸O, δ²H, and d-excess values show obvious seasonal variations, being depleted in winter and enriched in summer. Although temperature is an important variable that affects the change in δ¹⁸O values of precipitation, the results of the sub-cloud evaporation effect and moisture tracing show that differences in moisture sources caused by seasonal changes in large-scale water moisture transport are an important cause of seasonal changes in δ¹⁸O and d-excess of precipitation at NTP. Depleted δ¹⁸O and enriched d-excess in winter represent the source of moisture transported by the westerlies from the Mediterranean area and Central Asia. Enriched δ¹⁸O and d-excess values in summer precipitation are related to the temperature effect. In addition, the meridional motion of the atmospheric flow has an effect on the precipitation isotope values in the NTP. When the meridional circulation is enhanced, the water vapour from low latitudes is easily transported northwards, enriching the summer precipitation isotope values in the central and eastern parts of the plateau. This provides a new insight into the explanation of stable oxygen isotopes in climate proxies across the westerlies-dominated Tibetan Plateau.

1. Introduction

Atmospheric precipitation is a crucial component of the hydrological cycle and a source of land surface water [1,2]. The changes of atmospheric precipitation is closely linked to the management and utilization of local water resources, extreme natural disasters and ecological and environmental problems in arid and semi-arid areas [3,4]. Due to the different atomic binding energies and diffusion rates of water molecules with different masses, the heavier (lighter) water molecules condense (evaporate) preferentially during the phase transition process, resulting in isotopic changes in precipitation that record the physical processes that can occur from the moisture source region to the raindrop region, such as evaporation of raindrops under clouds, mixing of moisture from different sources, and changes in local environmental conditions [5,6,7,8]. Understanding the dominant role of these factors in spatiotemporal variations in precipitation isotopes is crucial for correctly characterizing climate signals preserved in natural archives such as ice cores [9,10], tree rings [11,12,13], and lake sediments [11,14,15,16,17].

In recent decades, great strides have been made in investigating the environmental significance of stable isotope signals from water bodies on the Tibetan Plateau (TP) [18]. There is now a comprehensive understanding of the spatial and temporal distribution, as well as the primary controlling factors, of atmospheric precipitation stable isotopes over the TP, with the identification of three distinct models. The westerly mode predominates in the northern TP, influenced by the westerly circulation. This results in δ¹⁸O depletion in winter and enrichment in summer, indicating a significant temperature effect throughout the year [19,20]. The monsoon mode mainly occurs in the southern TP, governed by the monsoon circulation. The δ¹⁸O of precipitation is depleted in summer and is manifested as the amount effect, but subsequent studies have confirmed that the depleted δ¹⁸O is associated with the precipitation process and convective activity in both the local and source area during the water vapor transport process [21,22,23]. On an annual scale, the influence of the strength of upstream convective activity on precipitation isotopes in the TP is regulated by Walker Circulation and Monsoon Circulation. During La Niña events, intensified convective activities over the tropical Indo-Western Pacific Ocean led to a decrease in the condensation temperature of cloud tops, resulting in the transport of depleted water vapor to the plateau with the Indian Summer Monsoon (ISM) [24,25,26]. The transition mode mainly occurs in the central TP, which is characterized by the alternating control of the westerlies and the monsoon circulation, and, combined with sub-cloud evaporation and local water vapour recycling, precipitation isotope change is complicated [27,28].

Seasonal variations in precipitation-stable isotopes are explicable through moisture sources and regional environmental conditions [29,30,31]. Dansgaard introduced the concept of “deuterium surplus” (d-excess = δD-8δ¹⁸O) to signify ocean moisture source conditions, primarily influenced by sea surface temperature, relative humidity, and wind speed. The global mean d-excess value is 10‰ [32,33]. Generally, the high (low) d-excess value indicates continental (oceanic) sources of water vapor [34,35]. The d-excess value gradually increases from south to north in the TP, suggesting a decreasing influence of oceanic moisture sources. The ISM affects the spatial pattern of precipitation in the TP through two paths, with one entering directly from the southwest border and the other turning southeast from the southwest border and entering the plateau from the Brahmaputra Valley [24,36,37]. The spatial differences in precipitation isotope variation during pre-monsoon, monsoon outbreak, and monsoon retreat on the TP serve as indicators of the monsoon development. The ISM first arrives in the TP in the southeast (mid to late June) and progressively moves northwestward. Conversely, the time for monsoon retreat gradually becomes later from north to south of the TP [23,38,39,40]. Nevertheless, using the d-excess value to characterize moisture sources in precipitation relies on the assumption that d-excess remains constant during moisture transport, which contradicts the actual situation [41,42,43,44]. During the moisture transport process, the sub-cloud evaporation effect decreases the d-excess value, while the recirculated moisture increases the d-excess value [45,46,47]. Research on the influence of moisture sources on precipitation isotopes in the TP areas has predominantly concentrated on the southern TP [35,48,49] and the northeastern and western TP [19,50]. However, there is also interest in the far north, such as the Tianshan Mountains [2,51,52,53], with relatively few investigations on moisture sources in the north of the TP, where sub-cloud evaporation and continental recycling processes hold significance. Consequently, it is imperative to elucidate the impact of moisture sources, transport processes, and local environmental factors on precipitation to understand the mechanism governing regional precipitation isotopes.

The Northern Tibetan Plateau (NTP) (33° N~45° N, 75° E~100° E) is located in the central part of the Eurasian continent, with a general elevation exceeding 4000 m. It falls within the arid and semi-arid zone and is characterized by limited ocean moisture and reliance on westerly moisture transport, glacial meltwater, and local moisture recycling [54,55,56]. Several studies have established a positive correlation between stable isotopes of precipitation and temperature in this region [8,18,20,57]. With global warming, the NTP tends to be warmer and wetter, and evapotranspiration in the TP shows an upward trend during 1970–2015, particularly in the eastern and southern regions, as opposed to the northwestern region [58]. Furthermore, an evaluation of sub-cloud evaporation based on GNIP site data from 1960 to 2018 has revealed an escalating trend in the TP [7]. These changes, coupled with the expansion of lakes, reduction of the glacier area, permafrost degradation, extreme weather events, and growth of vegetation cover, will inevitably modify the water cycle pattern and impact the spatial and temporal variation characteristics of precipitation isotopes [28,59,60,61]. To comprehensively understand the seasonal variation of moisture sources and the relationship between stable isotopes and regional environmental conditions in the NTP, we collected one-year precipitation samples from six stations to investigate the seasonal variation of precipitation isotopes in diverse regions. Our study involved analyzing the controlling factors of the seasonal variation of precipitation stable isotopes, evaluating the impact of sub-cloud evaporation, employing the HYSPLIT model to trace the moisture source of precipitation, and elucidating the reasons for the seasonal variation of precipitation isotopes in different regions, as well as quantifying the contribution of the influencing factors.

2. Study Area

The study area spans the Northern Tibetan Plateau (NTP) and Tianshan mountains, covering 10 degrees in latitude and 25 degrees in longitude (Figure 1). Most sampling sites are situated in the alpine climate zone, which is characterized by elevations exceeding 3000 m above sea level, with alpine desert-steppe or alpine meadow terrain. The average annual temperature remains below 0 °C, while the precipitation amount range from 115 to 440 mm, primarily concentrated from May to October. The Tianshan (TS) region stands as a “wet island” within the arid and semi-arid northwest China, receiving an average annual rainfall of over 400 mm due to westerly winds transporting moisture along the east-west Oriental mountain range [52,62]. The Muztagh (MZTG) site in the Pamir region of the northwestern TP and Qilianshan station (QLS) in the northeastern TP endure westerly winds throughout the year, featuring a cold climate with average wind speeds surpassing 4.5 m/s, however, MZTG experiences lower precipitation and relative humidity compared to Qilianshan (refer to Supplementary Materials Table S1) [51,63,64]. Ngari (NL), Beiluhe (BLH), and Sanjiangyuan (SJY), which stretch from the western to the eastern plateau, lie approximately on the same latitudinal belt. NL is situated in the Maga grasslands, 10 km from Pangong Co, which is characterized by a cold and dry climate with an annual rainfall of 118.65 mm and a relative humidity of 29.88%, respectively. BLH is in the plateau’s hinterland and a permanent permafrost area [65,66], experiencing an annual wind speed of 4.23 m/s. SJY lies on the Indian monsoon margin, receiving an average annual rainfall exceeding 420 mm and is influenced by moisture from the Arabian Sea and Bay of Bengal during strong monsoon periods [67,68]. These six stations cover a wide geographical area, exhibiting diverse climatic and environmental conditions, leading to variations in the performance of precipitation isotopes across the region.

3. Data and Methods

3.1. Precipitation Sampling and Isotopes

A total of 360 precipitation samples were collected from six different stations (MZTG, NL, TS, BLH, QLS, and SJY) located at various coordinates in the years 2018–2020 (see Table S1 in Supplementary Materials). The precipitation samples were manually collected using a 30 cm diameter polyethylene funnel and stored in a 5 L polyethylene collection bottle. Following each precipitation event, rain or snow samples were drained through the funnel at the bottom of pre-cleaned polyethylene bottles. The polyethylene bottles and funnels were arranged at the onset of each rain event, and the collected water samples were subsequently transferred to 30 mL polyethylene bottles and stored at 4 °C in a refrigerator.

The hydrogen and oxygen stable isotope values of the 360 precipitation samples were analyzed using the Picarro-L2140i wavelength scanning optical cavity ring-down spectrometer at the State Key Laboratory of Tibet Plateau Earth System, Environment and Resources, Institute of Tibetan Plateau Research, Chinese Academy of Sciences. To ensure accuracy and reliability, two calibrated methods were employed. (1) Each sample and isotopic standard underwent seven sequential injections, discarding the first three injections due to memory effects. The arithmetic average of the last three injections constituted the final measured value. (2) Samples and standard isotopes were injected sequentially two and seven times, respectively. One standard sample was injected for every six samples, and at the end of the sample sequence, three standard samples were injected in descending order of their values. The average of the two values was selected as the final value, expressed as δ-values relative to V-SMOW (Equation (1)). The measurement accuracy for δ¹⁸O and δD was 0.05‰ and 0.5‰, respectively.

$$\mathsf{\delta }=\frac{{\mathrm{R}}_{\mathrm{sample}}-{\mathrm{R}}_{\mathrm{standard}}}{{\mathrm{R}}_{\mathrm{standard}}}\times 1000‰$$

(1)

3.2. Evaluation of Influence from Sub-Cloud Evaporation on δ¹⁸O Composition

Sub-cloud evaporation denotes the alteration in precipitation isotopes due to evaporation as raindrops descend from the cloud base to the ground through unsaturated air [69,70]. Over the past five decades, the sub-cloud evaporation effect of precipitation isotopes has garnered considerable attention in the cold and arid regions of China, particularly in the TP [7]. It is a multifaceted physical process influenced by various factors, including the distance from the cloud base to the ground, raindrop size, and air temperature and humidity [47,71]. In this study, the Stewart model is utilized to quantify the intensity of the sub-cloud evaporation effect, with the calculations detailed below [70,72]:

$$∆d=d-d_{cloud}=2F-{81}^{8}F$$

(2)

$$\mathrm{ⁿ}F=\left(1-\frac{\mathrm{ⁿ}\gamma }{\mathrm{ⁿ}\alpha }\right)\left(f^{n\beta }-1\right)$$

(3)

$$\mathrm{ⁿ}\gamma =\frac{\mathrm{ⁿ}\alpha h}{1-\mathrm{ⁿ}\alpha \left(D/D^{\prime }\right){\mathrm{ⁿ}}^{\prime }(1-h)}$$

(4)

$$\mathrm{ⁿ}\beta =\frac{1-\mathrm{ⁿ}\alpha (D/D\prime ){\mathrm{ⁿ}}^{\prime }(1-h)}{\mathrm{ⁿ}\alpha (D/D\prime ){\mathrm{ⁿ}}^{\prime }(1-h)}$$

(5)

where ²F and ¹⁸F in Equations (2) and (3) are the changes of ²H and ¹⁸O from the cloud base to the ground, respectively. ²γ, ¹⁸γ, ²β, and ¹⁸β are calculated using Equations (4) and (5). The parameters ²D/²D′, ¹⁸D/¹⁸D′, and n′ are 1.024, 1.029, and 0.58, respectively [69,73,74]. h represents the relative humidity, and ⁿα represents the equilibrium fractionation factor under temperature control calculated using Equations (6) and (7), where T is the ambient temperature and the unit is K.

$${10}^3\ln {2\mathsf{\alpha }}^{+}=\frac{{1158.8T}^3}{{10}^9}-\frac{{1620.1T}^2}{{10}^6}+\frac{794.84T}{{10}^3}-161.04+\frac{{2.9992\times 10}^9}{T^3}$$

(6)

$${10}^3\ln {1^8\mathsf{\alpha }}^{+}=-7.685+\frac{6.7123{\times 10}^3}{T}-\frac{{1.6664\times 10}^6}{T^2}+\frac{{0.35041\times 10}^9}{T^3}$$

(7)

$f$ in Equation (3) is expressed as the ratio between the mass of the raindrops falling on the ground (m_g) and the sum of this variable with the mass of the evaporation loss below the cloud (m_ev) (Equation (8)). m_ev is equal to the product of the evaporation intensity (E) and the raindrop falling time (t) and is calculated using Equation (9). E is determined by two functions (A₁, A₂), where A₁ is a function of temperature (°C) and raindrop diameter (cm) and A₂ is a function of temperature (°C) and relative humidity (%). There are different A₁ and A₂ values under certain conditions, which can be obtained by bilinear interpolation [70].

$$f=\frac{m_g}{m_g+m_{ev}}$$

(8)

$$m_{ev}=Et$$

(9)

$$E=A_1\left(T,R\right)A_2(T,h)$$

(10)

The time taken for a raindrop to fall from the cloud base to the ground is determined by the height of the cloud base ($H_{c-g}$(km)), which can be obtained from empirical Equation (12) and the final velocity of the raindrop (V_drop(m/s)) [75].

Where T_ave is the mean air temperature of the column of air between the lifting condensation height (LCL) and the ground, the unit is K. S₀ and S_LCL are the pressures of the ground and the LCL, respectively. S_LCL is defined by Equation (13).

$$t=H_{c-g}/V_{drop}$$

(11)

$$H_{c-g}=\mathrm{18,400}\left(1+T_{ave}/273\right)\ln S_0/S_{LCL}$$

(12)

$$S_{LCL}=S_0{\left(T_{LCL}/T_0\right)}^{3.5}$$

(13)

$$T_{LCL}=T_{dp}-(0.001296T_{dp}+0.1963)(T_0-T_{dp})$$

(14)

where T₀ and T_LCL are the air temperatures at the ground and LCL and T_dp is the dew point temperature of the ground. T_LCL was calculated using the empirical Formula (14).

$$V_{drop}=\left\{\begin{array}{c}9.58e^{0.0354H_{c-g}}\left[1-e^{-{\left(R/1.77\right)}^{1.147}}\right],0.3\le R<6.0 \\ 1.88e^{0.0256H_{c-g}}\left[1-e^{-{\left(R/0.304\right)}^{1.819}}\right],0.05\le R<0.3\\ 28.40R^2{e}^{0.0172H_{c-g}},R<0.05 \end{array}\right.$$

(15)

V_drop (m/s) is related to the falling height of the raindrops and the diameter of the raindrops (cm), which can be obtained from Equation (15).

$$D_{50}={0.69}^{1/c}A{I}^q$$

(16)

c, A, and q are 2.25, 1.30, and 0.232, respectively, where D is the mean diameter of the raindrops calculated from Equation (16) and I is the rainfall intensity (mm·h⁻¹).

$$m_g=\frac{4}{3}\pi r_{end}^3\rho$$

(17)

Assuming that the raindrops falling on the ground are perfectly spherical, m_g is calculated from Equation (17), where r_end is the radius of the raindrop when it hits the ground and $\rho$ is the density of water.

3.3. Back Trajectory Calculation

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, developed by the Air Resources Laboratory at the National Oceanic and Atmospheric Administration, is used to identify the moisture path and air parcel trajectories using NCEP/NCAR reanalysis data. Previous studies have shown that moisture absorption on the track occurs in a short period of time (less than 7 days) [76,77], whereas the average residence time of water moisture in the atmosphere is 10 days [78,79]. To accommodate the dry and high evaporation climatic conditions in the study area, the tracking time was set to 7 days. Given that the altitude of the study area ranges between 2000 and 4000 m, in order to capture the water vapor transport tracks at different altitudes as much as possible, operational heights in the model were defined as 100, 500, 1000, 1500, 2000, and 2500 m. Finally, cluster analysis was employed to clearly discern the direction and influence of moisture sources on precipitation isotopes, and the number of track clusters was determined based on the Total Spatial Variance (TSV) change diagram. The modifications enhanced the clarity, corrected the punctuation, and added more concise explanations to the paragraph while maintaining the academic style and improving the readability of the content.

3.4. Meteorological Data and Reanalysis Datasets

The meteorological data are predominantly sourced from automatic weather stations located near the sampling points, providing hourly readings of temperature, precipitation, relative humidity, wind speed, and air pressure. Reanalysis data driving the HYSPLIT model are obtained from the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR) in GDAS format, spanning from January 1948 to the present with a horizontal resolution of 2.5° × 2.5°. The data can be downloaded from https://www.esrl.noaa.gov/ (20 December 2022). Additionally, the daily values of dew point temperature (T_dp), air column mean temperature (T_mean), specific humidity (q), meridional wind speed (u), and zonal wind speed (v) are extracted from the ERA-Interim reanalysis results of the European Centre for Medium-Range Weather Forecasts (ECMWF), which are available at https://www.ecmwf.int/en/forecasts/datasets/browse-reanalysis-datasets (20 December 2022), with a horizontal resolution of 0.25° × 0.25°.

4. Results

4.1. Seasonal Variability of Precipitation Isotopes

The annual data were categorized into seasons including spring (MAM), summer (JJA), autumn (SON), and winter (DJF), as well as the winter half-year (October–April) and the summer half-year (May–September) (see Table S2 in Supplementary Materials). The study reported a range of precipitation δ¹⁸O values from −31.67‰ to 7.42‰, and precipitation δ²H values ranging from −253.34‰ to 80.25‰. The d-excess values varied between −40.77‰ and 32.07‰, with a mean of 11.49‰.

Figure 2 illustrates the seasonal variations of precipitation stable isotopes (δ¹⁸O, δD, and d-excess) at various sampling sites. The δ¹⁸O values in MZTG, NL, TS, BLH, QLS, and SJY range from −31.67‰ to 7.42‰, with mean values of −12.02‰, −9.25‰, −10.64‰, −8.58‰, −7.54‰, and −12.58‰, respectively. The lowest mean annual value is recorded at −12.58‰ in SJY, while the maximum is −7.54‰ in QLS. The precipitation isotope values observed in our study area fall within the global range of precipitation isotope variations, which typically span between −50‰ and 10‰. Additionally, the seasonal variation of precipitation isotopes at six stations exhibits a consistent pattern and is characterized by continuous depletion in winter and enrichment in summer. This observation aligns with previous studies on precipitation patterns within the predominant influence area of the westerlies [18,46,80,81,82].

The d-excess values at MZTG, NL, TS, BLH, QLS, and SJY stations range from −40.77‰ to 32.07‰, with mean values of 10.46‰, 10.5‰, 9.25‰, 9.29‰, 17.68‰, and 12.66‰, respectively. These results are comparable to those observed in adjacent areas, such as the Qinghai Lake Basin [19], the western Pamir Mountains [53], the Tianshan Mountains [74,83], the Qilian Mountains [27], and the northwestern TP [73]. The seasonal fluctuation of d-excess, though smaller than that of δ¹⁸O, still follows a pattern of being lower in the winter half and higher in the summer half, except at SJY and QLS stations. Despite this general trend, the d-excess value does not align with that of the ocean evaporation source [84,85,86] and remains highly sensitive to local evaporation [28,53,72]. The maximum d-excess at most stations (except for MZTG) occurs in SON, lagging behind the season of the maximum value of δ¹⁸O in JJA. This suggests that the increased air moisture from local surface water evaporation, soil, and plant transpiration in SON compared to the strong evaporation season of JJA, which correlates with the high temperature, may account for this phenomenon. Notably, the more consistent seasonal d-excess variation at SJY indicates that precipitation processes at this station are distinctively influenced by specific moisture source conditions and atmospheric properties compared to the other stations.

4.2. Spatial Variability of Precipitation Isotopes

The mean δ¹⁸O values of TS (−17.78‰), MZTG (−26.14‰), NL (−22.3‰), BLH (−16.51‰), and SJY (−16.4‰) in the winter half show a decreasing variation from north to south and an increasing variation from west to east, which is attributed to the gradual precipitation process resulting from moisture transported over long distances by the westerly wind. The lower mean δ¹⁸O values in DJF at MZTG (−28.8‰) and BLH (−22.09‰) compared to other stations may be associated with lower condensation temperatures at higher altitudes. In contrast, the δ¹⁸O values at both stations show higher values in JJA (MZTG: −3.48‰; BLH: −7.33‰) due to a more localized moisture cycle with rising temperatures, and the strong evaporation leads to the depletion of moisture in surface water, entering the atmosphere and participating in precipitation. Gui et al. suggest that recirculating vapor contributes more to the atmosphere at higher altitudes due to the weak sub-cloud evaporative effect [27]. Additionally, the δ¹⁸O value of JJA in SJY (−10.9‰) located at the edge of the monsoon with the most depleted moisture of the six stations may be influenced by the monsoon carrying the marine air mass. Sun and Wang used the FLEXPART model to trace the source of moisture in the headwaters of the three rivers and found that moisture from the Bay of Bengal and Arabia, carried by the southwest monsoon, and from the Pacific, carried by the East Asian monsoon, can reach this region during JJA [68].

The d-excess values of TS (9.98‰), MZTG (13.78‰), NL (13.2‰), BLH (9.43‰), and SJY (12.5‰) in the summer half-year decrease in fluctuation from south to north and from west to east. MZTG exhibits the highest annual mean d-excess (10.46‰), while QLS (17.68‰) and MZTG (3.5‰) record the maximal values in the summer and winter half-years, respectively. These trends suggest a gradual increase in recycled moisture in the NTP from south to north and from west to east during the summer. In the Qilian Mountains, the local moisture circulation is enhanced in summer with increasing altitude [87], and the altitude effect on the d-excess is more pronounced in the summer (0.8‰/m) than in the winter (0.3‰/m) [27]. The lower d-excess value observed at MZTG in the winter half-year may be due to two factors: (1) the smaller number of samples collected in winter results in low d-excess values that represent individual precipitation events; (2) the moisture trajectory at MZTG indicates that 19% of the moisture in DJF originates from the Arctic Ocean (see Section 5.2). An available study in the western part of the Pamirs confirms that the d-excess value of polar moisture varies between 4‰ and 10‰ and fluctuates around 4‰ in DJF [83].

4.3. Local Meteoric Water Line

The linear relationship for δ¹⁸O and δD in precipitation is represented by the Global Meteoric Water Line (GMWL), which originally was expressed as δD = 8δ¹⁸O + 10 [88]. This was revised in 1993 to δD = 8.17δ¹⁸O + 10.35 [89]. The slope of the line reflects the degree of O and H fractionation and is highly sensitive to evaporation from moisture sources [33], sub-cloud evaporation of droplets falling through unsaturated air columns [90,91,92], and the mixing of multiple moisture sources [22]. Furthermore, the local meteoric water line (LMWL) is tailored to indicate the variation of environmental elements in distinct geographical areas.

Figure 3 shows the atmospheric precipitation line at different locations using regression analysis. There is a significant linear relationship between δD and δ¹⁸O.

• TS: δD = 7.97δ¹⁸O + 8.91 (R² = 0.98, N = 56)
• MZTG: δD = 8.41δ¹⁸O + 15.41 (R² = 0.99, N = 31)
• NL: δD = 8.59δ¹⁸O + 12.31 (R² = 0.93, N = 41)
• BLH:δD = 8.05δ¹⁸O + 9.74 (R² = 0.97, N = 94)
• QLS: δD = 8.51δ¹⁸O + 21.52 (R² = 0.98, N = 49)
• SJY: δD = 8.15δ¹⁸O + 14.52 (R² = 0.99, N = 87)

Compared with GMWL, the slope and intercept of LMWL in TS are slightly lower than those of GMWL, while the slope of LMWL in BLH and SJY are close to that of the GMWL, but the intercepts are significantly different (BLH: 9.74, SJY: 14.52). The TS exhibits a lower slope and intercept, while the MZTG, NL, and SJY demonstrate greater values than the GMWL. The decreased slope suggests that precipitation in TS may have undergone more sub-cloud evaporation of raindrops. In contrast, the higher altitude and lower temperatures at the other stations promote a weaker effect of sub-cloud evaporation, potentially contributing to their higher slope and intercept values, which could be attributed to moisture sources and local recycling. Nevertheless, the LMWL obtained from annual data does not capture the variation across different seasons due to seasonal changes in moisture sources, transport patterns, and local meteorological conditions. Therefore, we analyzed the seasonal variability of the precipitation line equation at different stations.

Changes in the slope and intercept at the six sites can be categorized into three types on a seasonal scale, as detailed in Table S3 in the Supplementary Materials: (1) the slope and intercept are less than eight and ten, respectively; (2) both the slope and intercept are greater than eight and ten, respectively; and (3) the slope is less than eight while the intercept is greater than ten. The first type is only observed in JJA and SON at TS and MZTG, suggesting the presence of a sub-cloud evaporation effect. The second type is predominantly found in DJF and JJA at SJY, QLS, and BLH, which is potentially attributed to minimal sub-cloud evaporation and heightened ambient humidity, and in MAM at MZTG due to amplified local continental recycling. The third type is observed in SON at SJY and QLS and in JJA at MZTG and QLS. When the temperature exceeds 0 °C, the intercept of the LMWL rises as a result of plant growth, transpiration, and evaporation from water bodies, suggesting an intensified local moisture cycle [93]. The slope of the LMWL may be lower in JJA and SON within arid and semi-arid regions.

4.4. Seasonal Variation of d-excess from the Cloud Base to the Ground (Δd)

The variability in d-excess between the cloud base and the ground (Δd) reflects the strength of sub-cloud evaporation. A lower Δd value indicates more intense sub-cloud evaporation. According to the Stewart model, the monthly averages of the d-excess between the cloud base and the ground from April to October at different sites are calculated in

Table 1. d-excess composition of precipitation at the level of the ground sampling site and the cloud base for 6 sampling sites over the north Tibetan Plateau (Δd represents the d-excess variability between the cloud base and the ground).

Location	d-excess	Apr	May	Jun	Jul	Aug	Sep	Oct	Monthly Mean
TS	Ground (‰)	9.05	14.79	7.49	3.42	11.12	10.88	10.59	9.62
	Cloud base (‰)	9.30	15.05	7.97	4.57	15.81	9.71	10.88	10.47
	Δd (‰)	0.25	0.26	0.48	1.15	4.69	−1.17	0.29	0.85
MZTG	Ground (‰)	-	16.73	22.48	13.74	20.78	-	-	18.43
	Cloud base (‰)	-	16.77	23.06	13.81	20.87	-	-	18.63
	Δd (‰)	-	0.04	0.58	0.07	0.09	-	-	0.20
NL	Ground (‰)	2.51	22.92	10.76	14.30	11.35	14.66	-	12.75
	Cloud base (‰)	5.98	23.86	18.09	15.48	12.44	16.20	-	15.34
	Δd (‰)	3.47	0.94	7.33	1.18	1.09	1.54		2.59
BLH	Ground (‰)	-	3.13	11.70	8.21	7.95	11.67	-	8.53
	Cloud base (‰)	-	31.49	18.50	15.22	14.54	16.46	-	19.24
	Δd (‰)	-	28.36	6.8	7.01	6.59	4.79	-	10.71
QLS	Ground (‰)	-	-	19.10	18.18	17.28	5.92	-	15.12
	Cloud base (‰)	-	-	19.44	18.69	17.58	6.00	-	15.43
	Δd (‰)	-	-	0.34	0.51	0.3	0.08	-	0.31
SJY	Ground (‰)	13.45	19.15	13.60	9.26	11.39	14.01	13.79	13.52
	Cloud base (‰)	21.60	19.61	14.93	10.00	11.80	16.51	15.61	15.72
	Δd (‰)	8.15	0.46	1.33	0.74	0.41	2.5	1.82	2.20

.

Annually, the d-excess values of cloud base precipitation at TS, MZTG, NL, BLH, QLS, and SJY stations are 10.47‰, 18.63‰, 15.34‰, 19.24‰, 15.43‰, and 15.72‰, while the corresponding values for ground precipitation are 9.62‰, 18.43‰, 12.75‰, 8.53‰, 15.12‰, and 13.52‰, respectively. The monthly mean △d values at TS, MZTG, NL, BLH, QLS, and SJY are 0.85‰, 0.195‰, 2.59‰, 10.71‰, 0.31‰, and 2.20‰, suggesting weaker sub-cloud evaporation at BLH, NL, and SJY compared to TS, MZTG, and QLS. Throughout April to October, the Δd values at the TS station are 0.25‰, 0.26‰, 0.48‰, 1.15‰, 4.69‰, −1.17‰, and 0.29‰ for each month. The results indicated that the sub-cloud evaporation effect was stronger in April, May, June, September, and October at TS situated in the canyon area of the upper reaches of the Urumqi River, compared to July and August.

Precipitation events with temperatures below 0 °C from October to March may exhibit negligible sub-cloud evaporation. The stations included in this paper experienced precipitation events with temperatures above 0 °C concentrated from May to October, resulting in varying monthly trends in Δd, precipitation amount, relative humidity, and temperature (Figure 4). The Δd values at MZTG from May to August are 0.04‰, 0.58‰, 0.07‰, and 0.09‰, whereas those at QLS from June to September are 0.34‰, 0.51‰, 0.3‰, and 0.08‰. In comparison to MZTG and QLS, NL exhibited a more positive seasonal variation in Δd, ranging from 0.94‰ in May to 7.33‰ in July. BLH showed Δd values of 28.36‰, 6.8‰, 7.01‰, 6.59‰, 4.79‰, and −1.17‰ in April, May, June, July, August, and September, consecutively. According to Figure 4, the intensified sub-cloud evaporation effect in late spring is linked to a temperature increase, whereas in summer, it is influenced by the combination of high temperature and low humidity.

Hence, the sub-cloud evaporation effect in late spring and early summer at the NTP sites is stronger than that in late summer and early autumn, differing significantly from the northwestern region of China [70,94] and the Loess Plateau [95], where the most pronounced sub-evaporation effect occurs in spring. Meanwhile, the spatial variation of the d-excess at the cloud base exceeds that of the d-excess at the ground. This suggests that the cloud bases of these stations may have varying moisture sources, highlighting that precipitation isotopes are influenced not only by the sub-evaporation of raindrops but also by the mixing of diverse moisture sources.

4.5. Correlation between the Remaining Fraction of Raindrops (f) and Change in Δd

The relationship between Δd and the remaining fraction of raindrops (f) has been utilized to estimate the raindrop evaporation rate based on the d-excess value in precipitation. Sub-cloud evaporation is a complex physical process influenced by temperature, humidity, and raindrop size. For instance, a 5 °C rise in temperature can decrease Δd by 0.3 to 4.0‰, while a 10% increase in relative humidity can raise Δd by 1.1 to 10.3‰ [70]. Previous studies have indicated that this relationship exhibits variation across different regions, particularly in arid and semi-arid regions with low precipitation events [95], resulting in an underestimation of precipitation isotope values [96,97].

The relationships between f and Δd at six stations in the NTP are depicted in Figure 5, demonstrating a notable correlation. The correlation coefficients between f and Δd from May to September at NL and MZTG, which are relatively dry, are less than 1, indicating that a 1% increase in f results in Δd increases by 0.48‰ and 0.31‰, respectively. QLS and BLH are situated at an altitude exceeding 4000 m, and the average temperature from May to September does not exceed 8 °C. The correlation coefficients between f and Δd are also less than 1, with a 1% increase in f leading to Δd increasing by 0.33‰ and 0.63‰, respectively. Under the most humid environmental conditions in TS and SJY, the correlation coefficients between f and Δd are close to 1‰/1%, which is consistent with the European Alps [72], the TS Mountains, and the Hexi Corridor [98]. Our study indicates that the linear relationship between f and Δd is stronger under environmental conditions of relatively high humidity, high precipitation, and low temperature. However, the linear relationship of 1‰/1% between f and Δd is not universally applicable in cold and dry regions, potentially being significantly lower than this empirical relationship value in reality.

5. Discussion

5.1. Local Meteorological Variables and Precipitation Isotopes

The temporal and spatial variations of isotopes in precipitation are generally influenced by meteorological parameters, including local temperature, precipitation, relative humidity, atmospheric pressure, and wind speed.

Table 2. Statistical correlation between precipitation δ¹⁸O and local meteorological parameters at sampling sites over the north Tibetan Plateau during different seasons.

Linear Correlation Coefficient
Sites	Time	δ18O-T(R2)	δ18O-RH(R2)	δ18O-Pre(R2)	d-excess-T(R2)	d-excess-RH(R2)	d-excess-Pre(R2)	d-excess-δ18O(R2)
TS	Annal	0.92 *(0.80)	0.08(0.01)	−0.06(0.14)	0.13(0.01)	−0.20(0.04)	0.06(0.11)	−0.03(0.001)
	Summer	0.99 *(0.59)	0.13(0.12)	−0.08 *(0.40)	−0.77(0.19)	−0.19(0.06)	0.10(0.28)	−0.43 *(0.34)
	Winter	0.71 *(0.66)	−0.08(0.02)	0.08 *(0.34)	0.57(0.17)	−0.25(0.03)	−0.01(0.01)	0.17(0.07)
MZTG	Annal	2.21 *(0.87)	−0.22(0.07)	0.53(0.55)	0.65(0.11)	0.04(0.003)	0.64(0.13)	0.39(0.12)
	Summer	2.02 *(0.70)	−0.05(0.05)	−0.79(0.26)	0.32(0.01)	0.10(0.02)	0.54(0.10)	0.10(0.02)
	Winter	1.64 *(0.66)	−0.01(0.21)	−0.33(0.07)	−0.60(0.11)	0.12 *(0.4)	0.85 *(0.61)	−0.91(0.66)
NL	Annal	0.34(0.10)	−0.13(0.03)	0.09(0.001)	1.72 *(0.42)	−0.07(0.003)	0.93(0.05)	0.10(0.06)
	Summer	0.02(0.0001)	−0.17(0.08)	−0.25(0.01)	−0.51(0.05)	−0.11(0.02)	0.16(0.005)	0.53 *(0.34)
	Winter	−0.64(0.15)	−0.01(0.001)	0.12(0.0001)	1.73(0.1)	−0.86(0.04)	−2.12(0.42)	−0.24 *(0.64)
BLH	Annal	0.38 *(0.13)	0.01(0.002)	-	−0.04(0.0008)	−0.008(0.0004)	-	0.03(0.001)
	Summer	−0.12(0.01)	−0.07(0.07)	-	−0.39(0.04)	−0.06(0.02)	-	−0.01(0.00001)
	Winter	0.35(0.07)	0.03(0.02)	-	1.03(0.19)	0.18(0.15)	-	0.09(0.024)
QLS	Annal	0.59(0.10)	−0.18 *(0.24)	−0.02(0.002)	0.52(0.04)	−0.08(0.02)	−0.14(0.07)	0.26(0.15)
	Summer	0.59(0.10)	−0.18 *(0.24)	−0.02(0.002)	0.52(0.04)	−0.08(0.02)	−0.14(0.07)	0.26(0.15)
	Winter	-	-	-	-	-	-	-
SJY	Annal	0.43 *(0.25)	0.03(0.003)	0.05(0.0002)	−0.10(0.01)	−0.17(0.09)	−0.62(0.02)	0.15(0.02)
	Summer	0.12(0.002)	−0.25 *(0.12)	−0.35(0.01)	−0.57(0.05)	−0.27 *(0.11)	−0.38(0.01)	0.17(0.04)
	Winter	0.56 *(0.30)	0.03(0.003)	−1.88(0.01)	−0.04(0.001)	−0.15(0.08)	−5.47(0.03)	0.15(0.03)

Note: T: temperature (°C); Pre: precipitation amount (mm); RH: relative humidity (%); numbers with * means the correlations are significant at the 0.05 level; numbers without an * do not pass the significant test.

illustrates the correlation between daily measurements of stable isotopes in precipitation and the meteorological parameters of temperature, precipitation, relative humidity, and wind speed at the six stations. On an annual scale, all six sites in our study demonstrate temperature effects, with slight variations in seasonal changes, which is consistent with previous findings. However, the correlation with relative humidity and precipitation did not achieve statistical significance. The δ¹⁸O of summer precipitation is negatively correlated with relative humidity only at QLS and SJY.

5.1.1. The Temperature Effect

Previous studies have confirmed that the temperature effect on precipitation isotopes is significant in the midlatitudes of the Northern Hemisphere [32,83,99,100,101]. In the western and northeastern areas of the Tibetan Plateau (TP) situated north of 30° N, the gradient in the relationship between temperature and precipitation oxygen isotope values exceeds 0.4‰/°C [102].

Except for NL, the other five stations exhibit a strong positive correlation between precipitation δ¹⁸O and temperature. Among them, the annual gradient of daily precipitation isotope and temperature at TS reaches 0.92‰/°C, surpassing the finding of Pang et al. (0.45‰/°C) [103], but resembling the results of Yao et al. (0.87‰/°C) [104]. The temperature gradient of the MZTG reaches 2.01‰/°C, exceeding that of global mid and high latitudes (0.55‰/°C) [89]. Precipitation isotopes in BLH and NL show a positive correlation with temperature on an annual scale, but they lack significance on a seasonal scale. SJY exhibits a temperature effect during the winter half-year and shows a negative correlation with relative humidity in JJA. QLS lacks winter samples, but δ¹⁸O in precipitation is positively correlated with summer temperature (0.59‰/°C), which is consistent with previous research [27].

To better quantify the impact of temperature effects on precipitation isotopes at each station, all precipitation samples are categorized into three groups based on temperature: (Ⅰ) T < 0 °C; (Ⅱ) 0 °C <T< 10 °C; and (Ⅲ) T >10 °C, and the findings are depicted in Figure 6. In range I, the precipitation isotopes of these stations display a significant positive correlation with temperature, except for NL, with the temperature effect being most notable at MZTG and TS. In range II, precipitation isotope values are unlikely to change with increasing temperature because the enriched isotope caused by the temperature effect is offset by the depleted isotope caused by the recirculated moisture from evaporation and transpiration. A striking feature of range III is that the δ¹⁸O increases and the d-excess decreases with increasing temperature, indicating the existence of a sub-cloud evaporation effect. This is supported by the negative correlation between the gradient of d-excess/T in this range (see Table S4 in Supplementary Materials).

5.1.2. The Precipitation Amount Effect

Most of the NTP stations receive relatively low precipitation, leading to a weak impact of precipitation amount on the precipitation isotope fraction. In our study, the correlations between precipitation isotopes and precipitation amount at the six stations do not yield statistically significant results at either the annual or seasonal scale. Previous studies have indicated that the influence of the precipitation amount is primarily evident in the monsoon region (south of 30° N) and the transition region (30~35° N) during the summer period [24,28,36,48]. Of the six stations included in this study, NL, BLH, and SJY are in the transition zone near the ISM boundary. They are affected by the winter westerlies, which tend to show a ‘temperature effect’ but demonstrate an ‘amount effect’ when there is an increase in rainfall under the enhanced conditions of the summer monsoon. Table 2 illustrates a negative correlation between precipitation isotopes and relative humidity during summer at SJY (r = −0.25, p < 0.05) and QLS (r = −0.15, p < 0.05), indicating that the moisture carried by the monsoon contributes to a more humid regional environment.

5.2. Backward Trajectory and Precipitation Isotopes

The Indian Summer Monsoon (ISM) and the westerlies are the two major atmospheric currents that influence the TP, resulting in precipitation-stable isotopes showing three domains (e.g., the westerlies domain, the monsoon domain, and the transition domain) [18]. In general, δ¹⁸O in precipitation associated with the westerlies show depleted characteristics due to long-distance precipitation processes and low temperatures, while d-excess is more enriched compared to monsoon-derived moisture. Nevertheless, multiple evaporation–condensation processes and local meteorological conditions during the transfer process may obscure the characteristic signal of the moisture source, causing regional precipitation isotopes to exhibit unique characteristics.

The HYSPLY model is employed to reconstruct the movement paths of moisture during precipitation events at six stations in NTP. The trajectories were clustered at various altitudes based on the seasons. In the case of NL, MZTG, and QLS, there are few samples in DJF, and these trajectories are clustered in MAM, JJA, and SON. The corresponding results are presented in Figure 7 and Figure 8.

5.2.1. Stations Mainly Dominated by the Westerlies

The primary moisture sources of TS and MZTG are transported by the westerlies from northern Europe (NE), the Mediterranean Sea (MS), and the Arabian Peninsula (AP), as well as from the Arctic Ocean (AO) and neighboring central Asia (CA). During the DJF at TS, moisture originated from the westerlies, with contributions from NE (31%), MS (12%), and the Arabian Sea (AS) (19%), while moisture from AO and CA accounted for 13% and 26%, respectively. The average δ¹⁸O and d-excess in precipitation were −25.44‰ and 0.01‰, respectively. During MAM, the moisture transported from NE, MS, and AS gradually decreased from 62% to 46%, while the moisture from CA increased to 45%. Consequently, the δ¹⁸O and d-excess values gradually enriched to −14.39‰ and 10.86‰, respectively. During JJA, 16% of the moisture originated from the AO, 34% from the westerly pathway, and 50% from CA, leading to an enrichment of δ¹⁸O in precipitation to −4.72‰, while the d-excess decreased to 8.65‰. In SON, 8% of the moisture originated from the MS, 15% from the North Atlantic (NA), and 77% from CA, resulting in a decrease in precipitation δ¹⁸O to −10.95‰, while the d-excess was 12.01‰ due to increased moisture from the surrounding area.

The moisture source of MZTG is similar to that of TS, with distinct seasonal variations. During MAM, moisture was transported from NE (8%), the NA (10%), MS (16%), CA (45%), and the polar area (19%). Nearly half of the westerly path’s moisture originated from CA. The precipitation values were −21.1‰ for δ¹⁸O and 6.76‰ for d-excess. During JJA and SON, the composition of moisture sources remained largely similar to MAM, with an increased proportion of moisture from CA. In DJF, the NA exhibits high surface relative humidity and low temperature, resulting in a low d-excess value of the transported moisture due to the weak dynamic fractionation effect.

At the local scale, the precipitation d-excess value at TS decreased in JJA due to rising temperatures and the sub-cloud evaporation effect. The seasonal variation of precipitation isotopes and moisture source at MZTG was consistent with that at TS, and the mean annual d-excess closely resembled that of Wuqia (12.9‰) in the western Tianshan Mountains. However, during the winter half-year, the d-excess value was only 3.5‰ at MZTG, in contrast to 8.02‰ at TS. This discrepancy was attributed to 19% of the precipitation at MZTG originating from polar moisture (PM) in MAM, indicated by trajectory clustering analysis, with the d-excess value of PM ranging from 4‰ to 5‰ [105]. During the summer half-year, the d-excess value was 9.98‰ at TS, contrasting with the 13.78‰ at MZTG. During JJA at MZTG, 49% of the moisture came from CA, and 19% was from the Xinjiang region, resulting in a high d-excess value attributed to local recirculation.

5.2.2. Stations in the Transition Zone between the Westerlies and India Summer Monsoon

NL and BLH are situated in a transition zone influenced by both the ISM and the westerlies. In MAM, the westerlies brought moisture to NL from the NA, MS, AP, and the adjacent continental areas, accounting for 17%, 28%, 14%, and 41%, respectively, with mean δ¹⁸O and d-excess values in precipitation of −18.6‰ and 12.7‰, respectively. During JJA, the proportions of moisture from the Indian Ocean (IO), North Indian Peninsula (NIP), and Northwest China (NWCA) were 5%, 46%, and 31%, respectively, resulting in enriched mean δ¹⁸O values of −7.09‰, while d-excess remained stable at 12.4‰. In SON, moisture proportions from the Bay of Bengal (BOB), CA, and the westerly path were 10%, 61%, and 29%, respectively. The precipitation moisture from the IO exhibited depleted isotopic characteristics, exemplified by the δ¹⁸O values of 19 August and 30 September, which were significantly lower than the monthly JJA (−7.09‰) and SON (−9.58‰) averages.

In DJF at BLH, westerlies transported 37% of the total precipitation from NA, MS, and AS, with moisture from neighboring continents and the south of TP constituting 16% and 47%, respectively. The mean values of δ¹⁸O and d-excess in precipitation were −22.09‰ and 6.29‰, respectively. During MAM, the moisture transport path at BLH remained mostly stable, with the proportion from the neighboring mainland increasing to 53%, including 12% of the moisture from the NIP. The mean values of δ¹⁸O and d-excess in precipitation were −11.94‰ and 4.62‰, respectively. The increase in the mean δ¹⁸O of precipitation from −22.0 ‰ in DJF to −11.94 ‰ in MAM could not be solely attributed to the temperature effect. Therefore, the higher δ¹⁸O in precipitation might be linked to continental moisture. In addition to seasonal variation in moisture sources, the higher mean δ¹⁸O values at NL and BLH in JJA and SON compared to DJF and MAM could be attributed to enhanced local recycling and the sub-cloud evaporation effect.

The primary sources of moisture for SJY and QLS include the NA, MS, AP, and CA transported by the westerlies, and moisture from NIP and the BOB carried by ISM. Moisture from the MS and AP is linked to higher d-excess values in precipitation, while moisture from NIP and BOB is associated with lower d-excess values. Conversely, the d-excess value in CA is typically lower than that from MS due to multiple evaporation and condensation processes during transportation. During DJF, the primary moisture sources at SJY included NA (18%), MS (18%), AP (28%), and NWCA (36%). The mean values of δ¹⁸O and d-excess in precipitation were −18.2‰ and 13.8‰, respectively. During MAM, a reduced amount of moisture is transported by the westerlies, principally from MS and AP. The remaining 43% of precipitation moisture originates from NWCA, with mean δ¹⁸O and d-excess values of −11.9‰ and 13.4‰, respectively. During JJA, under the influence of monsoon activity, 4% and 8% of the SJY precipitation’s moisture originated from NIP and BOB, respectively. The proportion of moisture from NWCA decreased to 27%, with 48% of the moisture coming from Inner Mongolia (IM) in northern China, likely due to the influence of the East Asian monsoon. The mean values of δ¹⁸O and d-excess in precipitation were −9.89‰ and 11.6‰, respectively. During SON, the westerlies’ intensity increased and the monsoon influence gradually weakened; however, moisture from NWCA remained dominant. The mean values of δ¹⁸O and d-excess in precipitation were −16.4‰ and 14‰, respectively, during this period.

During MAM, the MS, CA, and PM contributed 6%, 61%, and 33% of the moisture at QLS, respectively. The average values of δ¹⁸O and d-excess in precipitation were −11.36‰ and 11.78‰, respectively. During JJA, the proportion of moisture transported by PM decreased, while the contribution from NWCA increased, resulting in a mean δ¹⁸O and d-excess in precipitation of −6.9‰ and 18.08‰, respectively. During SON, the proportion of moisture from NA increased from 9% to 25%, with the majority of the remaining moisture originating from CA, NWCA, and IM, resulting in a mean δ¹⁸O and d-excess in precipitation of −9.89‰ and 17.99‰, respectively. The trend of d-excess variation in precipitation in SJY during MAM (13.4‰), JJA (11.6‰), and SON (14‰) was at odds with that of QLS during MAM (11.78‰), JJA (18.08‰), and SON (17.99‰), suggesting differing mechanisms at play in different seasons. In September, the enriched δ¹⁸O and d-excess values were primarily influenced by a greater contribution of local recirculated moisture, while the depleted δ¹⁸O and d-excess in SON resulted from mixed moisture originating from IM. The fluctuation trend of d-excess at the cloud base in SJY gradually decreased from April to July but increased from July to October. This pattern aligns with the seasonal variation of the moisture source, transitioning from westerlies in DJF to enhanced local continental recycling from MAM to JJA.

5.3. Moisture Flux and Wind Field with Precipitation Isotopes

The moisture flux represents the intensity of moisture transport, which is defined as the mass of moisture flowing through a unit area per unit of time. Typically, it is combined with the wind field to pinpoint the moisture source. The trajectory of the air mass gives a broad indication of the water source direction rather than an exact determination. Utilizing ERA5 reanalysis data, specific humidity, meridional, and zonal wind speeds were used to calculate the moisture flux at 850 hPa for October, January, April, and July of 2018 to 2020.

Figure 9 and Figure 10 illustrate that the westerlies and the ISM predominantly governed the moisture transport, displaying varying spatial patterns across seasons. From October to April, a high-pressure ridge at 80–90°E over the TP predominated, causing the westerlies to carry cold and dry air, leading to decreased moisture content during DJF. The prevailing Indian low-pressure systems from July to October led to the ISM bringing warm moisture from the southeastern edge of the TP into the plateau’s interior. The blocking effect of the Tanggula Mountains restricted the influx of moisture into the plateau’s interior, resulting in amplified rainfall across the entire plateau. Additionally, a small amount of rainfall was brought by the East Asian monsoon in October, extending to the eastern part of the plateau.

Seasonal variations in moisture flux demonstrate the impact of westerly intensity and the ISM on the isotopic composition of precipitation at NTP. This study conducts a correlation analysis between the isotopic data of precipitation from May to September at NL, BLH, and SJY, and the meridional and zonal wind speeds at 850 hPa, 500 hPa, and 300 hPa over the study area from 1948 to 2020, to investigate the influence of the meridional position and the strength of the westerlies on the precipitation isotopes. This analysis is depicted in Figure 11. The isotopic composition of summer precipitation at NL and BLH exhibits a negative correlation with the meridional wind speed, with this correlation gradually weakening with increasing distance and altitude. Notably, it weakly shows a positive correlation with the meridional wind speed at the 300 hPa level in BLH. Strengthening of the meridional circulation leads to a depletion of precipitation isotopes at NL and BLH. Moving from west to east, the correlation between summer precipitation isotopes and meridional wind speed at NL, BLH, and SJY transitions from negative to positive, with this positive correlation being particularly significant in the southeastern and eastern regions of the TP. Enhanced meridional circulation facilitates the northward transportation of moisture from the low-latitude Indian Ocean, consequently influencing the isotopic composition of precipitation in the SJY.

We conducted a correlation analysis between precipitation isotopes and zonal wind speeds at the three stations during the summer season, as shown in Figure 12. The findings reveal that precipitation isotopes at NL exhibit a weak negative correlation with zonal wind speed, whereas those at BLH and SJY demonstrate a distinct positive correlation with zonal wind speed. The notable positive correlation between precipitation isotopes and zonal wind within the 25–30° N range suggests that the strengthening westerlies drive the eastward movement of the southerly wind, consequently leading to a higher enrichment of precipitation isotopes in summer compared to DJF at BLH and SJY.

6. Conclusions

In this study, we analyzed the seasonal variations of the precipitation isotope values at six stations in the NTP and highlighted the impact of the sub-cloud evaporation effect and moisture source on the δ¹⁸O values in precipitation. The main local meteorological factor affecting the isotopic signature of precipitation is air temperature, which is consistent with previous findings about temperature effects in this region, but the negative correlation between precipitation δ¹⁸O and relative humidity at SJY and QLS indicates that moisture brought by the summer monsoon is the reason for the depletion of δ¹⁸O compared to other stations. In addition, the sub-cloud evaporation effect is widespread in the NTP, and the effect is more obvious in late spring and early summer, but the linear correlation coefficient between f and Δd is found to be close to 1‰/% only at SJY and TS, which is not consistently observed at other cold and arid stations. Therefore, simply applying the slope of 1‰/% to describe the evaporation process under the cloud may result in significant errors, which should be approached with caution in practical applications.

Local meteorological factors can only explain part of the seasonal variation of precipitation isotopes; the moisture source tracing indicates that variations in large-scale moisture transport significantly contribute to seasonal fluctuations in δ¹⁸O and d-excess levels in NTP’s precipitation. During winter, the depletion of δ¹⁸O and the enrichment of d-excess signify the long-distance transportation of moisture by the westerly winds, and the depletion of d-excess is linked to Arctic moisture. Continental water sources, including those from central Asia and Northwest China, contribute significantly to annual precipitation events and bring more enriched moisture compared to oceanic sources from the Atlantic and Arctic Oceans. Moisture sources at various stations during JJA are intricate; the enrichment of δ¹⁸O and d-excess values in precipitation is linked to the temperature effect and the reinforcement of local continental recycling.

Interestingly, we found that the precipitation isotopes exhibit a negative correlation with meridional wind speed, which gradually weakens with greater distance and altitude at NL and BLH. Moving from west to east, the correlation between summer precipitation isotopes and meridional wind speed shifts from negative to positive at NL, BLH, and SJY. This suggests that the meridional shift of the westerly winds impacts the precipitation isotope values in the NTP. With the strengthening of meridional circulation, enriched moisture from low latitudes is readily transported northwards, thereby enriching the isotopic values of summer precipitation in the central and eastern plateau areas. This provides an up-to-date reference for the explanation of stable oxygen isotopes in climate proxies across the westerlies-dominated Tibetan Plateau.