A detailed study of the climatic significance of δ18O in precipitation was completed on a 1500 km southwest-northeast transect of the Tibetan Plateau in central Asia. Precipitation samples were collected at four meteorological stations for up to 9 years. This study shows that the gradual impact of monsoon precipitation affects the spatial variation of δ18O-T relationship along the transect. Strong monsoon activity in the southern Tibetan Plateau results in high precipitation rates and more depleted heavy isotopes. This depletion mechanism is described as a precipitation “amount effect” and results in a poor δ18O-T relationship at both seasonal and annual scales. In the middle of the Tibetan Plateau, the effects of the monsoon are diminished but continue to cause a reduced correlation of δ18O and temperature at the annual scale. At the monthly scale, however, a significant δ18O-T relationship does exist. To the north of the Tibetan Plateau beyond the extent of the effects of monsoon precipitation, δ18O in precipitation shows a strong temperature dependence. δ18O records from two shallow ice cores and historic air temperature data were compared to verify the modern δ18O-T relationship. δ18O in Dunde ice core was positively correlated with air temperature from a nearby meteorological station in the north of the plateau. The δ18O variation in an ice core from the southern Plateau, however, was inversely correlated with precipitation amount at a nearby meteorological station and also the accumulation record in the ice core. The long-term variation of δ18O in the ice core record in the monsoon regions of the southern Tibetan Plateau suggest past monsoon seasons were probably more expansive. It is still unclear, however, how changes in large-scale atmosphere circulation might influence summer monsoon precipitation on the Tibetan Plateau.
 The Tibetan Plateau, often described as the world's third pole, is the largest and highest plateau in the world. Evidence of a warming global climate and the possibility of decreasing summer monsoon precipitation may have significant impact on the extent of glacial ice on the Tibetan Plateau. Thus it is of interest to reconstruct past climatic conditions on the Tibetan Plateau to determine if this region has been exposed to similar historic warming trends providing insight to the region's response to changing climatic conditions.
 Although stable isotopes are widely used in various fields, such as hydrology, meteorology, atmospheric physics, ecology, one of its fundamental uses is in paleoclimate study. This can be attributed to the recognition of empirical relation between δ18O in recent precipitation and observed air temperature [Dansgaard, 1964; Rozanski et al., 1993]. Thus, the δ18O record in ice cores provides evidence of long-term climate change. A global network was established to monitor the long-term relationship between stable isotopes in precipitation and temperature in 1960s. However, these stations are not equally distributed, and expansion of the network to areas near various climate archives and at high altitude is suggested [Araguas-Araguas et al., 2000]. On the Tibetan Plateau there have been limited studies of stable isotopes in precipitation, especially for continuous long-term observations [Araguas-Araguas et al., 1998]. In addition, climate modeling efforts on the Tibetan Plateau are difficult because the large spatial differences cannot be defined due to the lack of empirical observations on the Tibetan Plateau [Cole et al., 1994].
 The complexity of moisture cycling on the Tibetan Plateau has an apparent impact on δ18O fractionation in precipitation. Thus, the spatial variability of δ18O in precipitation across the Tibetan Plateau is a major concern when attempting to describe δ18O-temperature relationships. Despite these potential problems, the wide distribution and high elevation of glaciers on the Tibetan Plateau qualify this region as ideal for low and midlatitude alpine ice core studies. Several ice cores have been recovered from various areas on the Plateau. Three ice cores were recovered from the Dunde ice cap in the Qilian Mountains on the northeastern Tibetan Plateau in 1987 [Thompson et al., 1989; Yao and Thompson, 1992]. The Dasuopu ice cores were recovered from the Himalayas on the southern Tibetan Plateau [Thompson et al., 2000].
 The variation of δ18O in recent precipitation on the Plateau and surrounding area was studied in an attempt to establish a δ18O-temperature relationship and reconstruct past air temperature records using δ18O records from ice cores. Araguas-Araguas et al.  describe the large-scale variation of stable isotopes in precipitation in East Asia. In-situ observations of local scale were also performed in various areas on and around the Plateau. [Yao et al., 1996, 1999; Zhang et al., 1995; Aizen et al., 1996]. Earlier studies have established the δ18O-temperature relation in the northeast of the plateau near the Dunde ice core, and also in the Urumqi River basin in Tianshan Mountains [Yao et al., 1996, 1999].
 The stable isotope composition in precipitation is related to precipitation characteristics such as the moisture source region and the degree of rainout [Rozanski et al., 1993; Schoch-Fisher et al., 1984; Hubner et al., 1979]. A southwest-northeast transect of the Plateau along the moisture trajectory provides an opportunity to study how moisture movement affects the climatic significance of δ18O in precipitation. A network monitoring stable isotopes in precipitation was launched in 1991 on the northeastern Plateau and extended to other areas in the following years. These stations are geographically close to the ice core sites, and are at high elevation. Thus it is assumed that δ18O in recent precipitation in close proximity to the ice-coring sites are comparable to δ18O measured in the ice cores. The scope of this work includes the collection of precipitation samples at four meteorological stations from 1991 to 1999. The purpose of this work is to evaluate the climatic significance of δ18O in recent precipitation and shallow ice core records by establishing a δ18O-temperature relationship while considering variations in precipitation source and amount. The approach is to correlate δ18O in recent precipitation events to air temperatures recorded during the event. The δ18O-temperature relationship will then be applied to δ18O values from ice cores to reconstruct past temperature variations.
2. Methods and Materials
2.1. Study Area
 The study area extends from the southern Tibetan Plateau, dominated by the monsoon climate, to the north of the Tibetan Plateau where continental air masses dominate (Figure 1). Precipitation was sampled from south to north at 4 meteorological stations. Nyalam is in the southern-most site located where the precipitation source is predominantly monsoon. The climate in this region is humid with an annual precipitation of about 650 mm (Table 1). Compared to the other stations, Nyalam has significantly more precipitation in winter and a comparatively small annual air temperature amplitude (Figure 2). Lhasa is located in a vast valley near the Yalongzhangbo River (named Brahmaptra after entering India). This site is also dominated by monsoon precipitation. The main moisture trajectory moves from the south along the Brahmaptra toward the southern Tibetan Plateau [Lin and Wu, 1990; Gao et al., 1985]. Most of the precipitation occurs as rainfall from May to September (accounting for about 85% of total annual precipitation). Very little precipitation occurs as snow (4% of annual precipitation from November to March) at this location.
Table 1. The Spatial Variation of δ18O and Air Temperature at Four Meteorological Stations on the Tibetan Plateau
The seasonality is a ratio of precipitation from May to October to the whole year.
 Two more sampling sites are located to the north of the Tanggula Mountains, where continental air masses dominate, resulting in elevated δ18O in precipitation. Tuotuohe has an annual precipitation of about 260 mm. It has the greatest elevation of the four sites, which is reflected in the comparatively low annual air temperature. Delingha is located on the northeast edge of the Tibetan Plateau. This site is completely dominated by continental air masses with the highest annual amplitude of air temperature variation and the lowest annual precipitation (Table 1; Figure 2).
2.2. Precipitation Sampling
 All precipitation samples were event-based. Rainfall samples were collected and sealed in plastic bottles immediately following an event to limit evaporation. Snow and other solid precipitation samples were collected and melted at room temperature and then processed the same as rainfall samples. Samples were taken back periodically and stored frozen in the laboratory until analysis. δ18O was measured on all precipitation samples by a MAT-252 mass spectrometer (precision of 0.2‰) in the Laboratory of Ice Core and Cold Regions Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Lanzhou, China. The observation duration at each station is also listed in Table 1.
2.3. Meteorological Data
 Air temperature, precipitation amount and other meteorological data were recorded right before and after each precipitation event. These data were used to compare with δ18O in precipitation. The meteorological monthly air temperature and precipitation data at Nyalam (1985–1998) and Qilian Tuole (1957–1987) were also used in this study to compare with the ice core δ18O records.
2.4. Ice Core Records
 δ18O records from two shallow ice cores were compared to δ18O records in recent precipitation. δ18O was measured in the top 16 meters of an ice core recovered from the top of the Dunde ice cap in the Qilian Mountains on the northeastern Tibetan Plateau (38°06′N, 96°25′E, 5325 meters a.s.l.) in 1987 (Figure 1). This ice core was dated using methods described in Yao et al. . The upper 16 meters of the core, spanning about 30 years, was analyzed for δ18O. Air temperature data from a nearby meteorological station (Qilian Tuole) were used to develop the δ18O-temperature relationship. Another ice core was recovered from the Dasuopu Glacier (28°06′N, 85°46′E, 7000 meters a.s.l.) in the middle of Himalayas in 1996 [Duan et al., 2000] on the southern Plateau.
3. Results and Discussion
3.1. Temporal Fluctuation of δ18O in Precipitation
 From south to north, the seasonal pattern of δ18O does show a continuous change. A distinct seasonal difference of δ18O variations can be observed between the northern two stations (Delingha and Tuotuohe) and the southern two stations (Lhasa and Nyalam) (Figure 3). δ18O values at Delingha reflect continental moisture sources with typical δ18O seasonal variation. Similar δ18O variation can be found at Tuotuohe, but with a wider range of δ18O in the summer. At Lhasa and Nyalam, high summer temperatures are coincident with the lowest δ18O in precipitation. At Nyalam, the highest δ18O values are in the spring just before the monsoon season (March to May) and are coincident with low precipitation months.
 Compared with Nyalam, the lower δ18O at Lhasa, are probably due to a stronger monsoon rainout process along the moisture trajectories. The Toutuohe site is dominated by a continental climate, however, a wider range of δ18O values in the summer compared to the Delingha site suggest a slight impact of monsoon precipitation. δ18O in precipitation at the Delingha site is typical of a continental climate (no influence of monsoon precipitation), with expected seasonal variation.
3.2. Temperature Dependence of δ18O in Precipitation
 Several factors may affect the relationship between δ18O in precipitation and temperature. The local spatial variation of δ18O may be a factor at sites exposed to homogenous moisture origin. Naftz et al. , however, showed this variation to be negligible on a glacier in western North America. In the monsoon region, however, the moisture source variation makes it impossible to establish a spatial relation between δ18O and air temperature. Table 1 lists the average δ18O and air temperature during the observation period as well as general meteorological conditions. Lhasa has the highest annual air temperature and the lowest annual δ18O. In the northern Plateau annual δ18O is higher and annual air temperature is lower compared to the southern Plateau. The gradual increase of annual δ18O in the interior of the continent is a result of moisture recycling in the local area, which is not well understood [Araguas-Araguas et al., 1998]. Most precipitation at Toutuohe falls in solid form, even during the summer, therefore the fractionation due to the reevaporation of raindrops is not the cause of the highest δ18O values. The spatial variability of deuterium excess also confirms these observations [Tian et al., 2001b].
 Up to nine years of δ18O measurements in precipitation are not sufficient to establish a climate-scale relation between δ18O and air temperature. Thus, seasonal and event-based relationships between δ18O and air temperature are discussed.
 The significance of a linear regression least squares fit of δ18O in recent precipitation and air temperature increases from southwest to northeast on the Tibetan Plateau. A regression of the annual data sets for the Delingha location show a close relationship (R2 = 0.58) between δ18O in precipitation and air temperature (Figure 4). The results are given in equation (1):
Using the annual data sets, the significance of the δ18O–temperature relationships declines from northeast to southwest until no relation is observed at Lhasa and Nyalam.
Figure 5 shows the relationship between averaged monthly δ18O and air temperature at Delingha and Tuotuohe. Results from a linear regression least squares fit are given in equations (2) and (3):
It is clear that monthly averages of δ18O in precipitation and air temperature on the northern Plateau produce the best least squares fit and the strongest δ18O-temperature relationship. The spatial variation (north to south) of the δ18O-temperature relationship can be largely attributed to the seasonal variation of δ18O and air temperature as shown in Figure 6. The decrease of monthly δ18O compared with high air temperatures in the summer results from monsoon precipitation. From south to north, with the weakening of monsoon precipitation impact, this relation is increased. At Nyalam and Lhasa, δ18O is comparatively low due to summer monsoon precipitation, while summer temperatures remain high. Further north at Tuotuohe, which is more affected by continental air masses, a slight decrease in δ18O from the monsoon precipitation can also be observed from July to September, reducing the correlation between δ18O and temperature.
 The event-based relationship between δ18O and temperature can more precisely reflect the actual response of stable isotopes to temperature than the monthly relationship. This is for two main reasons: (1) the monthly variation of δ18O is possibly caused by changes in moisture sources rather than the temperature change, and (2) the large variation of δ18O in one month is averaged as in the case of Tuotuohe. At Delingha, however, there is consistent agreement between the event-based and monthly relationships. Thus, it appears that δ18O in precipitation responds to air temperature changes expectedly on the northern Plateau. At Tuotuohe, although there is a clear monthly relationship between δ18O and air temperature, the event-based relationship is poor. Clearly, both temperature and monsoon precipitation are interacting factors controlling δ18O toward the middle of the Plateau.
3.3. Precipitation Amount and δ18O in Precipitation
 Previous work has discussed the precipitation “amount effect” in monsoon precipitation [Dasngaard, 1964; Wei and Lin, 1994; Araguas-Araguas et al., 1995], but the mechanisms controlling the “amount effect” in monsoon are still unclear. There is no observed relation between δ18O and recent precipitation amounts at the northern sampling locations (Delingha and Toutuohe) and the multiregressions between δ18O and air temperature and recent precipitation amount did not improve correlation results in the earlier section. This suggests that precipitation amount has little impact on δ18O in recent precipitation in this region. In Figure 7, the monthly precipitation is compared with weighted δ18O in precipitation at the four stations on the Plateau. There is a general inverse correlation between monthly δ18O and precipitation amount at Nyalam and Lhasa (monsoon region).
 The two southern stations show clear precipitation “amount effect.” At Nyalam, there is a consistent seasonal variation of δ18O from high winter and spring values to very low summer values for multiyears. δ18O decreases dramatically during June and drops to the lowest values during July. This δ18O-temperature relation is inverse to what is expected in a continental climate. A winter snow storm, however, can result in extremely low δ18O values at Nyalam, as in December of 1997. The similarity of the monthly δ18O between the two southern stations indicates monsoon precipitation is a common moisture source and the phenomenon of “amount effect,” to a large extend, is related to large scale atmosphere circulation.
 The monsoon movement from south to north is most likely responsible for the gradual depletion of δ18O from south to north on the Plateau. From Katmandu (27°42′N, 85°22′E, 1340 m a.s.l.) to Nagqu (31°22′N, 91°32′E, 4530 m a.s.l.), the daily δ18O in precipitation in 1998 shows the lag of the temporal variation and the gradual depletion of δ18O along the monsoon trajectory from south to north, reflecting moisture movement in monsoon region (Figure 8). Further north at Tuotuohe, there is a slight “amount effect” during the summer, whereas there is no “amount effect” at Delingha. Examining the annual variation of δ18O and precipitation at Lhasa shows an obvious precipitation “amount effect” suggesting the same mechanisms control the seasonal and annual variation of stable isotopes in precipitation in monsoon regions. An arithmetic average of annual δ18O at Lhasa precipitation shows an inverse relation with annual precipitation (Figure 9).
 Thus, on the southern Plateau, the mechanism controlling the stable isotope in precipitation is closely related to the monsoon moisture transport. A detailed study of the effects of monsoon activity on measured δ18O in daily precipitation does show an apparent inverse relationship between δ18O and monsoon activities [Tian et al., 2001a]. The strong monsoon activities bring increased precipitation to the southern Tibetan Plateau and low δ18O in precipitation. Thus the variation of δ18O in monsoon precipitation is probably related to the vapor source and vapor trajectory, which is still not clearly understood.
3.4. Comparison Between δ18O Record in Recent Precipitation and Ice Cores
 Ice core records from the Tibetan Plateau extend back to the Last Interglacial Age [Yao et al., 1997]. Stable isotopes of the Greenland ice are significantly correlated with temperature of the coast sites [White et al., 1997]. Here, short- term observations are used to interpret the long-term climate change in ice core record. An initial approach is to examine the correlations between δ18O in the short-term record against the ice core records. The longest monitoring period is more than three decades at only a few meteorological stations on the global scale [Rozanski et al., 1992]. There are no decade-scale records of δ18O on the Tibetan Plateau, but the meteorological stations have air temperature data dating back several decades on the Tibetan Plateau. Comparing the air temperature data and δ18O ice core record provides an alternative approach to assess climate change on the Tibetan Plateau during the past four decades.
Figure 10 is a comparison of the annual variation between air temperature at the Qilian Tuole meteorological station and the δ18O record from the Dunde ice core. Positive correlations are apparent in the upper core where ice dating is confirmed and in the lower core where the highest β activity at a 13 meter depth indicates a date of 1963. There are apparent disagreements, however, between δ18O and annual air temperature (1969–1975). One reason is probably related to dating error. A more plausible reason is that the precipitation events are sporadic while the air temperature is continuous record. A similar comparison between the δ18O record at Colle Gnigetti and Fiescherhorn and nearby instrumental temperature records also shows some qualitative climatic features from δ18O records [Rozanski et al., 1997].
 In the monsoon precipitation region of the southern Tibetan Plateau, the Dasuopu δ18O ice core record is compared with air temperature data from the nearby Nyalam meteorological station. This station is about 3800 meters a.s.l., and about 20 km away form the ice core drilling sites. Moreover, it is located on the south slope of the Himalayas and probably has the same moisture origins as in the Dasuopu glacier.
 The annual precipitation at Nyalam is correlated with variations in the accumulation record in Dasuopu ice core (Figure 11a), indicating the two records can be compared. The precipitation record at Nyalam meteorological station shows a clear inverse relation to the δ18O record in the shallow Dasuopu ice core (Figure 11b). Thompson  compared the annual δ18O and annual accumulation in the upper part of another Dasuopu ice core, spanning four decades, and also found an inverse relation.
 Although there appears to be no δ18O-temperature relationship in the monsoon region of the Tibetan Plateau (Figure 4), consistent results from recent precipitation and shallow ice core isotope records indicate there is an inverse relation between δ18O in the Dasuopu ice core and precipitation amount (“amount effect”) Moreover, monsoon precipitation appears to affect the δ18O variation in precipitation and ice cores at seasonal and annual variation and decadal timescales. Most likely, strong monsoon activity results in high precipitation rates and depletes the heavy isotopes by fractionation processes. Longer records are needed in the southern Tibetan Plateau to establish a δ18O-temperature relation that considers the effects of monsoon activity on δ18O in precipitation.
4. Summary and Conclusions
 Stable isotopes in precipitation are dependent on many factors from the origins of moisture, transport of moisture, moisture recycling and precipitation processes. The combination of these processes can result in complex isotope fractionation, making it difficult to understand isotope variation in natural environments. Here, we discuss the temporal and spatial patterns of δ18O in recent precipitation from south to north of the Tibetan Plateau with focus on the climatic significance of δ18O variation in ice cores based on recent precipitation observations. Other studies have shown that the variation of δ18O in precipitation is controlled by large-scale moisture movement (monsoon) on the Tibetan Plateau. From southern regions influenced by monsoon precipitation to the northern regions, influenced by continental air masses, the behavior of δ18O changes dramatically.
 In northern Tibetan Plateau (Delingha), there is a clear correlation between air temperature and δ18O in precipitation at both event-based and monthly scales. The annual slope of the δ18O-temperature linear regression (0.65) is similar to the annual slope (0.67) measured at the Greenland ice sheet [Johnsen et al., 1989]. The strong relation between δ18O and temperature most likely reflects a homogenous moisture source in the northern Tibetan Plateau.
 In the middle of the Tibetan Plateau (Toutuohe), the behavior of δ18O is somewhat complex owning to influence by both continental air mass and monsoon precipitation. At Toutuohe, there is a clear monthly relation between δ18O and air temperature. The event-based relation, however, is poor and can not reflect the response of δ18O to air temperature change.
 In southern Tibetan Plateau (Lhasa and Nyalam), monsoon activity has an overwhelming impact on the temporal and spatial variation of δ18O in precipitation resulting in very low δ18O in summer precipitation. The high δ18O values in winter and spring precipitation are most likely related to different moisture sources [Aizen et al., 1996]. A possible explanation for low δ18O in monsoon precipitation is the phenomenon referred to as “amount effect.” The “amount effect” can overwhelm the temperature effect and greatly diminish the δ18O-temperature relationship, especially at the annual variation scale.
 Several decades of air temperature data from two meteorological stations were used to assess climatic response of δ18O in two shallow ice cores. In agreement with the strong recent precipitation δ18O-temperature relationship, the northern Dunde ice core δ18O record also shows a correlation with a nearby meteorological air temperature record. There is no clear relation of δ18O in recent precipitation to air temperature on the southern Plateau. There is, however, an inverse correlation of δ18O to precipitation amount. The δ18O record from southern Dasuopu ice core also shows an inverse relation with the precipitation amount from a nearby meteorology station.
 It is difficult, however, to use short-term observations to explain the long-term ice core δ18O record in monsoon region as shown by Thompson et al. . Thompson's work indicates the decadal to century variation of the δ18O record in tropical ice core records reflects temperature variation. Yao et al.  indicate the relationship between δ18O and accumulation measured in the ice cores is dependent on timescale. At a decadal scale, there appears to be a positive relationship between accumulation and δ18O, while an inverse relationship is apparent at the century timescale. This suggests that different mechanisms may be controlling large-scale atmospheric circulation, and thus δ18O variation, at different timescales.
 Long-term changes of the continental air temperature may affect the atmospheric circulation and thus the moisture origin, trajectory and expansion of monsoon precipitation. A lake-core record suggests the variation of the sources of monsoon precipitation in western Tibetan Plateau [Gasse et al., 1991]. Additional ice core records in Tibetan Plateau also suggest that the summer monsoon extended beyond its present limit to reach the north and west of the Plateau [Thompson et al., 2000]. Moreover, an earlier precipitation study also showed significant spatial differences in δ18O as a result of moisture originating from different air masses and trajectories on the Tibetan Plateau [Tian et al., 2001b].
 Here, a δ18O-temperature relationship was established for the northern Tibetan Plateau and may be used to assess past temperature variations using δ18O records from ice cores in this region. On the southern Plateau, however, additional work is needed to compare the δ18O records in different ice cores in the monsoon region to identify possible long-term consistent variations, confirming trends in climatic variation, albeit not quantifiable, in regions influenced by the monsoon on the Tibetan Plateau. To understand the isotope record in monsoon ice cores, both the short-term and long-term mechanism controlling isotope in precipitation should be clarified. General Circulation Models, coupled with isotopic fractionation in hydrological cycle [Joussaume et al., 1984; Rozanski et al., 1997; Hoffmann and Haimann, 1997; Hoffmann et al., 1998; Jouzel et al., 1987], probably can provide insight into how different mechanisms play their roles in different timescales in monsoon area.
 This work is supported by the Collective Innovation of National Natural Science Foundation of China (40121101), Ministry of Science and Technology of People's Republic of China (2001CCB00300), National Natural Science Foundation of China (40271025), and GAME/Tibet project. We thank Zhang Xinping for the initial field work, Li Peiji for his great help, and Toshio Koike from Tokyo University for his consistent support since 1998 for this research. We also thank Sun Wenzhen for sample analysis in the laboratory. Special thanks are given to the two anonymous reviewers for very helpful suggestions. Our thanks are also given to all those involved with precipitation sample collection, without which this study would not be possible.