1 Introduction

Stable isotopes (2H, 18O) are components of natural water body and are very sensitive to environmental changes (Dansgaard 1953; Craig 1961). They record the internal information of water cycle process and are an effective method to track complex hydrological and climatic processes (Gat 1996; Clark and Fritz 1997; Edwards et al. 2010; Good et al. 2015). The stable isotope composition of precipitation not only provides a means for the study of modern processes of global and regional climate and hydrology but also is applied to quantitative paleoclimate reconstruction (Hoffmann and Heimann 1997; Johnson and Ingram 2004; Daley et al. 2012; Fernandoy et al. 2011). All of these studies require more information on stable isotopes in precipitation.

At present, the global network for isotope in precipitation (GNIP) has more than 1000 monitoring sites, but only 31 stations are included in China, and the development of large-scale isotope monitoring network is relatively slow. Based on the experience of GNIP and other countries in establishing isotope observation networks, China has successively established a series of precipitation isotope monitoring networks, such as the Chinese Network of Isotopes in Precipitation (CHNIP) (Song et al. 2007) and the Tibetan Plateau Network of Isotopes in Precipitation (TNIP) (Yao et al. 2013). Scholars have applied the precipitation isotope data of these sites to carry out a lot of basic research and applied research, which has promoted the development of isotope theory. In the 1950s, Dansgaard (1953) discovered that isotopes in precipitation gradually depleted as the condensation temperature decreased, which opened up the application of isotopes in the field of hydrological cycles. Subsequently, many scholars began to explore the factors affecting the variation of precipitation isotope values (Craig 1961; Gat 1980; Liu et al. 2009a, b; Zhu et al. 2016). The atmospheric circulation model was used to study the variation characteristics of large-scale precipitation stable isotopes (Tindall et al. 2009; Sjolte et al. 2011; Zhang et al. 2012; Samuels-Crow et al. 2014). The relationship between precipitation isotopes and climatic factors was discussed and the sources of precipitation water vapor was analyzed (Pang et al. 2005; Sengupta and Sarkar 2006; Yamanaka et al. 2007; Li et al. 2014). In addition, a large number of studies have been carried out in the Qinghai–Tibet Plateau (Tian et al. 2001; Yao et al. 2013), the southwest monsoon region (He et al. 2000; Pu et al. 2013), the southeast monsoon region and the northwest arid region (Liu et al. 2008, 2009a; Huang et al. 2015; Wang et al. 2018), as well as on the basin scale (Wu et al. 2011; Zhao et al. 2011; Li et al. 2013a; Pang et al. 2015; Zhu et al. 2019) in China.

The research of precipitation isotopes in China has been fruitful on the regional scale, but for a long time, it has been difficult to observe in some areas (high latitude and high altitude) on the national scale and the data sharing of existing observation stations is insufficient. Therefore, it is necessary to establish the spatial distribution pattern of precipitation isotopes in China according to the modeling methods of Bowen and Wilkinson (2002, hereinafter referred to as BW model) to make up for the precipitation isotopes blanks of the lacks observation areas. The BW model has been used in previous research and is worthy of reference (Bowen and Wilkinson 2002; Dutton et al. 2005; Liu et al. 2009b), but further studies expanding the available data and exploring the implications of precipitation isotope variability in China are needed. In this study, the annual mean δ18O data of 177 sites (including 99 observation sites and 8 ice core sampling sites) was used to (1) re-establish the spatial distribution pattern of precipitation isotopes across China and (2) invert the atmospheric hydrological process and analyze the influencing factors of precipitation isotope changes in different climate regions of China. This study improves the accuracy of spatial distribution of precipitation isotopes over China and provides a new platform for understanding the spatial variability and influencing factors of precipitation stable isotope in China.

2 Data and method

2.1 Data sources

A total of 117 δ18O data including precipitation observation sites and ice core sampling points were collected. Figure 1 shows the spatial distribution of these sites in China. Among them, the δ18O data of 31 sites is from the GNIP dataset (http://www-naweb.iaea.org), our research team provides the measured δ18O data of 3 sites, and the remaining 83 sites are from publicly published research results in domestic and international journals. Detailed data information for each site is listed in Table 1. The precipitation isotope data provided by GNIP, the technical procedures for sample collection and transportation, and the standard data are strictly formulated by the International Atomic Energy Agency (IAEA). The isotope analysis was performed by the laboratory of the institute of cooperation of member states through sample comparison, and all analytical errors were controlled. In addition to GNIP sites, Chinese scholars and research institutions have established precipitation sample collection sites nationwide to measure isotope data by referring to GNIP sampling and experimental procedures. The error of the precipitation samples of each site is the measurement error of the analyzer itself. When selecting these sites, the measurement error of the precipitation δ18O value is controlled to be between ± 0.1 and ± 0.5‰. All of the above measured oxygen isotope data is expressed as δ values relative to V-SMOW (Vienna Standard Mean Ocean Water):

$$ \updelta {}^{18}\mathrm{O}=\frac{R_{\mathrm{sample}}-{R}_{\mathrm{standard}}}{R_{\mathrm{standard}}}\times 1000{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}}, $$
(1)

where Rsample is the ratio of 18O/16O in the samples and Rstandard is the ratio of 18O/16O in V-SMOW.

Fig. 1
figure 1

Spatial distribution of precipitation observation sites and ice core sampling sites and different climatic zones (the southeastern monsoon region can be divided into the southern region and the northern region with the boundary of the Qinling-Huaihe River) in China

Full size image
Table 1 Basic information of precipitation observation sites and ice core sampling sites and their δ18O value in China
Full size table

The δ18O value of precipitation is the weighted mean of precipitation amount for one or more years, calculated as follows:

$$ {\delta}_{(w)}=\sum {P}_i{\delta}_i/\sum {P}_i, $$
(2)

where Pi is the precipitation amount and δi is the oxygen isotope value .

In addition, the long-term climatic data were derived from the WorldClim-Global Climate Data version 2 (http://www.world-clim.org) (Fick and Hijmans 2017) and are also listed in Table 1. The vapor flux was calculated using NCEP/NCAR 2.5° × 2.5° reanalysis data (NCEP/NCAR Reanalysis 1), which were obtained from the National Center for Environmental Prediction/National Center for Atmospheric Research (http://www.esrl.noaa.gov).

2.2 BW model

Bowen and Wilkinson (2002) established a spatial distribution model of precipitation oxygen isotope (referred to as the BW model), which is one of the most effective methods for estimating the spatial distribution of stable isotopes in precipitation (Bowen and Revenaugh 2003). The BW model considers that the isotopic composition of precipitation as the sum of temperature driven rainout effect and regional patterns of water vapor source and transport (Bowen and Revenaugh 2003). Usually, the temperature is mainly affected by latitude and altitude, so the BW model uses altitude and latitude as model parameters to simulate the δ18O value (Liu et al. 2009b). The BW model is expressed as follows:

$$ \delta {}^{18}O=a{\left|\mathrm{Lat}\right|}^2+b\mid \mathrm{Lat}\mid +c\mathrm{Alt}+d, $$
(3)

where δ18O is the preliminary estimated annual mean precipitation oxygen isotope value. Lat and Alt is the latitude (°) and altitude (m) of observation site, respectively. a, b, and c are empirical parameters; d is an intercept. The error of the BW model was assessed using mean bias error (MBE), mean absolute error (MAE), and root mean square error (RMSE). According to the BW model (Bowen and Wilkinson 2002; Bowen and Revenaugh 2003), the final spatial distribution pattern of precipitation isotope is a combination of model estimates and interpolated residual error using inverse distance weighted method in Arcgis 10.2 software.

3 Results and discussion

3.1 Basic pattern

It can be seen from the method that the BW model is a second-order polynomial regression model of δ18O and latitude and altitude. During the process of water vapor transport from low latitude to high latitude, water vapor condensation produces precipitation due to the gradual decrease of temperature, which makes the residual water vapor and the isotopes in the subsequent precipitation gradually depleted, showing the latitude effect of the change of isotope value. In addition to the latitude effect, with the increase of altitude, the isotope value also shows a decreasing trend, that is, the elevation effect of the change of isotope value. In order to make the model effect more ideal, the BW model adopts a two-step regression method, which avoids the interaction of latitude effect and elevation effect. In order to minimize the effects of height, based on previous research experience (Bowen and Wilkinson 2002; Dutton et al. 2005) and the distribution of Chinese sites (Fig. 1), the first fit is the latitude and δ18O values of the site elevation below 200 m (n = 31). A second-order polynomial best describes the dependence of δ18O value on latitude:

$$ \delta {}^{18}\mathrm{O}=-0.009{\mathrm{Lat}}^2+0.427\mathrm{Lat}-11.323\kern0.24em \left({r}^2=0.92,P<0.0001\right), $$
(4)

It can be seen from Eq. (4) that the goodness of fit between δ18O and latitude is high, and there is a significant negative correlation between them. Figure 2a also shows that as the latitude increases, the δ18O value decreases gradually. Therefore, the isotope values of sites below 200 m over China show better latitude effect. The latitude effect is more obvious to the north of 30 °N (Fig. 2a).

Fig. 2
figure 2

a The fit of the δ18O value and latitude of the sites (n = 31) that altitude below 200 m. b The fit of the residuals and elevations of all sites (n = 117); the residual is the difference between the measured δ18O and the predicted δ18O by Eq. (4)

Full size image

The large elevation difference of China’s topography is an advantageous condition for studying the isotope elevation effect. In this study, the altitude of all sites ranged from 3 to 7010 m. In order to quantitatively estimate the change of δ18O value with altitude, first, the δ18O values of all sites were predicted according to Eq. (4) and then the residuals of each station were calculated. Second, the residuals and elevations of each site were fitted by least squares regression:

$$ \mathrm{RES}=-0.0015\mathrm{Alt}\kern0.24em \left({r}^2=0.65,P<0.0001\right), $$
(5)

where RES is residual. The residual is the difference between the measured δ18O and the modeled δ18O by Eq. (4). RES = measured δ18O-modeled δ18O; modeled δ18O was predicted by Eq. (4). As shown in Fig. 2b, the precipitation δ18O value in China shows a linear decline with altitude, and the slope of Eq. (5) is − 0.0015. That is to say, the precipitation δ18O value in China decreases with altitude by − 1.5‰ km−1. This value is lower than that of the global average of − 2.8‰ km−1 (Poage and Chamberlain 2001) and − 2‰ km−1 (Bowen and Wilkinson 2002), but it is consistent with previous studies on China (Liu et al. 2009b; Yang et al. 2014).

The types of landforms in China are complex and diverse, and the variation of δ18O value is influenced by many factors at the same time, such as topographic changes, local water vapor circulation, and sub-cloud secondary evaporation, which makes the decline rate of the δ18O value in precipitation with altitude vary greatly in time and space. Therefore, the variation of the simulated precipitation δ18O value with altitude in China actually simplifies the complexity of these influencing factors.

Based on BW model, the dependence of precipitation δ18O value on latitude and altitude combination is described by adding the above Eqs. (4) and (5). Equation (6) provides a good estimate of the δ18O value in precipitation (r2 = 0.64, P < 0.0001, Fig. 3a) in the following:

$$ \delta {}^{18}\mathrm{O}=-0.009{\mathrm{Lat}}^2+0.427\mathrm{Lat}-0.0015\mathrm{Alt}-11.323\kern0.24em \left({r}^2=0.64,P<0.0001\right), $$
(6)
Fig. 3
figure 3

a Measured δ18O value versus δ18O value modeled by Eq. (6) of 117 sites. b The frequency distribution of residual between measured δ18O value and modeled δ18O value

Full size image

As shown in Fig. 3b, the range of residual varies from − 4.59 to 5.3‰. The mean bias error (MBE), mean absolute error (MAE), and root mean square error (RMSE) of the modeled values are − 0.5‰, 1.5‰, and 2.09‰, respectively.

3.2 Spatial distribution of model residual

The variation range of the residual between the measured δ18O value and the modeled δ18O value is − 4.6 to 5.2‰ (Fig. 3b). As shown in Fig. 4a, the distribution of residuals is mostly between − 1.5 and 1.5‰, which further indicates that the BW model can better estimate the spatial distribution of δ18O value in precipitation over China. The spatial distribution of residual can reflect the influence of other important factors on the variation of precipitation isotope value in different regions except latitude effect and altitude effect (Eq. 6), such as water vapor source and transport, local water vapor cycle, sub-cloud secondary evaporation, precipitation amount effect, and relative humidity (Dutton et al. 2005).

Fig. 4
figure 4

a Spatial distribution of residuals using inverse-distance-weighted method over China. b Spatial distribution of δ18O value in precipitation over China based on a sum of residual distribution map (a) and δ18O value estimated by Eq. (6)

Full size image

In order to reflect the influence of the above regional factors on the precipitation δ18O value in different regions of China, we use the inverse distance weight (IDW) interpolation method to generate the spatial distribution map of the model 6 residuals (Fig. 4a). From Fig. 4a, we can see that the residuals are obviously larger in other regions except for the region of − 1.5 to 1.5‰, which indicates that the measured δ18O values in these regions are significantly different from the simulated δ18O values. Negative residual regions mainly occur in the southern Qinghai–Tibet Plateau, most of the Yunnan–Guizhou Plateau, and the Sichuan Basin; the south of the North Plain extends southward to the middle and lower reaches of the Yangtze river basin, the north of Taiwan, the southern slope of the Altai Mountains, and the Sanjiang Plain in Northeast China. These areas may also be affected by precipitation amount effect or Indian Ocean water vapor transport (Zhang and Yao 1998; Yao et al. 2013), making the measured of precipitation δ18O value lower than the estimated δ18O value. On the contrary, in the western and central regions of northwest arid region, such as the Taklimakan Desert in the Tarim Basin, Badain Jilin Desert in Inner Mongolia, and the north and west of Qinghai–Tibet Plateau, these areas may be affected by local continental water vapor and secondary evaporation (Tian et al. 2007; Liu et al. 2008), making the simulated δ18O value lower than the measured δ18O value.

3.3 Spatial distribution of δ18O value in precipitation over China

In order to reflect the combined effects of various factors on the precipitation δ18O value, the residual grid (Fig. 4a) was added to the grid generated by Eq. (6) to produce the spatial distribution map of precipitation δ18O value over China (Fig. 4b). Figure 4b represents our best estimation of precipitation δ18O value and can roughly show the spatial distribution of precipitation δ18O value over China.

It can be seen from the Fig. 4b that the high value region of δ18O is mainly distributed in the eastern monsoon region and the northwest arid region, and the low value area is distributed in the southwest of Qinghai–Tibet Plateau region and the northeastern region of China. In the eastern monsoon region, including most of the North China Plain, the Han River basin, the Sichuan basin, the Yunnan–Guizhou Plateau, the Yarlung Zangbo River basin, and the plains areas of Taiwan Island, the δ18O value is highest, which ranged from − 8 to − 6.3‰. In the western part of the Qinghai–Tibet Plateau and the northwest–southeast direction curved mountain, Altai Mountain, Tianshan, and other mountains, the δ18O value is lowest, which is between − 18.8 and − 16‰. Comparing our research results with other studies (Liu et al. 2009b; Yang et al. 2014), it is found that the range of the maximum δ18O value becomes smaller, and the range of the minimum δ18O value becomes larger, that is, the range of variation of δ18O becomes smaller, but the δ18O value is higher in most areas, especially in the central basin and the eastern low latitude plain. Obviously, the high value areas of δ18O are mainly distributed in basins, valleys, and low-altitude plains. The low-value areas are mainly distributed in high-latitudes and high-altitude areas and are strongly influenced by latitude effect and elevation effect.

In the eastern monsoon region, the δ18O value tends to decrease with the increase of latitude; when in the northeast region, the δ18O value is the lowest. Although the isotope values of precipitation in the monsoon region are generally higher, but the latitude effect can still be exhibited. The Qinghai–Tibet Plateau is affected by high altitude and special topography. Except for the plateau marginal zone, the Qaidam Basin, and the eastern low-altitude zone, the δ18O value of the plateau is low, and its spatial distribution is greatly affected by the northwest–southeast arc mountain. In the Northwest arid regions China, except for the low δ18O value in the mountainous area, the high value area of δ18O extends from the Tarim Basin to the Hexi area of Gansu and the Middle East area of Inner Mongolia, mainly affected by local water vapor cycle and local evaporation under arid environment.

3.4 Temperature and precipitation amount dependence of δ18O in precipitation

China’s temperature and precipitation vary greatly from north to south and from east to west. Therefore, China is divided into several sub-regions according to the relative uniformity of the underlying surface and climatic conditions (Fig. 1). Figure 5 shows the regression of δ18O values and temperature and precipitation amount for each sub-region. In the southeast monsoon region, although it can be affected by the southeast monsoon water vapor, there are great differences between the southern region and northern region. In the southern region, as the temperature rises and the precipitation increases, the δ18O value is positively correlated with the temperature and negatively correlated with the precipitation amount (Fig. 5a, b), that is, there are both temperature effect and precipitation amount effect. However, the temperature effect in the southern region is more significant than the precipitation amount effect and is closest to the subtropical ocean water vapor source. The δ18O value in the precipitation water vapor is high (Pang et al. 2005), and the high-value region of precipitation δ18O value is formed in the southern region with good hydrothermal combination (Fig. 4b). In the northern region, the δ18O value is positively correlated with temperature and precipitation (Fig. 5c, d), that is, there is a temperature effect but no precipitation amount effect, indicating that the precipitation amount effect is more obvious in low latitude area and the temperature effect is more obvious in middle- and high-latitude areas, which are more consistent with the previous scholars’ conclusions (Zhang and Yao 1998; Yamanaka et al. 2007).

Fig. 5
figure 5

The variation of δ18O values in the southern region (a, b) and northern region (c, d), northwest arid region (e, f), southwest monsoon region (g, h), and Qinghai–Tibet Plateau (i, j) with temperature and precipitation in China

Full size image

In the northwest arid area, the temperature effect is considered to be the decisive factor in many factors affecting the precipitation isotope composition, and the δ18O value in precipitation water vapor is originally high due to the influence of local continental water vapor cycle and secondary evaporation. As shown in Fig. 5e, f, there is indeed a temperature effect and no precipitation effect in the northwest arid area, but the correlation of the δ18O value with temperature and precipitation is weaker than that in other regions. This may be due to the sparse and uneven distribution of sites in the northwest arid region (Fig. 1), which makes the existing sites unable to represent the overall situation of the region. There is a positive correlation between the δ18O value and temperature and precipitation in the southwest monsoon region, and the correlation is high (Fig. 5g, h). In the Qinghai–Tibet Plateau region, there are still temperature effect and precipitation effect, but the correlation between δ18O value and temperature and precipitation is also weak (Fig. 5i, j), which is related to the special role of high terrain.

3.5 Water source signatures reflected by δ18O in precipitation

Figure 6a shows the annual average vapor flux, from which we can see the water vapor source and transport of China. Figure 6b shows the water vapor flux divergence, which represents the net income and expenditure of water vapor. In the eastern monsoon region, the westerly water vapor transport, the southwest monsoon, and southeast monsoon water vapor transport are converged in the south of the Yangtze River and the conveying intensity can reach 1200–1500 kg m−1 s−1. In the south and east of the convergence area, the water vapor mainly comes from the southwestern monsoon of the Indian Ocean and the southeastern monsoon of the Pacific Ocean. In the west and north of the convergence area, the water vapor is mainly transported by the westerly. Water vapor diffuses northward and eastward from the convergence area, forming a positive diffusion area in southeastern China, and the northward water vapor transport is weakened until the northernmost Mohe is only 300 kg m−1 s−1. The corresponding δ18O value in Fig. 4b also shows a decreasing trend with the increase of latitude.

Fig. 6
figure 6

Annual average vapor flux on vertically accumulated from 1981 to 2010 (a). Water vapor flux divergence (b). (Unit: kg m−2 s−1)

Full size image

The Indian Ocean water vapor is the main source of precipitation in the southern of the Qinghai–Tibet Plateau (Tian et al. 2001). In the west of China, the southwest monsoon water vapor transport from south to north is weakening, and the proportion of westerly water vapor and local continental air mass is gradually increasing (Fig. 6a). According to Tian et al. (2001), Tanggula Mountain is an important boundary. As shown in Fig. 4b, the precipitation δ18O value in western China decreases first and then increases from south to north. Figure 6b shows that in the south of the Loess Plateau and the north of Sichuan basin is the water vapor flux negative divergence area, i.e., the net income area of water vapor. In the south and southeast of the Qinghai–Tibet Plateau, due to the influence of high topography, water vapor around the flow and overturned transport to form the northwest–southeast water vapor convergence zone, which is one of the sources of precipitation water vapor in these areas, and the corresponding δ18O value is also larger. Therefore, the spatial distribution of precipitation δ18O values can well track water vapor and its transport trajectories from different sources.

4 Conclusion

A model was built to simulate the δ18O value in precipitation by stepwise regression of δ18O value and latitude and altitude of 117 sites in China. The spatial distribution map of precipitation δ18O value in China with high precision was generated by spatial interpolation, and the variation of δ18O value in different climatic regions and its influencing factors was analyzed. The simulation results show that the BW model can better simulate precipitation δ18O value in China. The main area of the model residual distribution in China is between − 1.5 and 1.5‰. Other residual deviation regions can be used to explain that the precipitation δ18O values are affected by other local factors besides latitude effect and altitude effect, while the spatial distribution of δ18O values over China can simultaneously reflect latitude effect, altitude effect, and regions other influence factors. The δ18O value is restricted by topography; the high value areas are mainly distributed in basins, valleys, and plains at low altitude; and the low value areas are mainly distributed in high latitude and high altitude areas and some mountain areas, which are affected by elevation effect. In the southeast monsoon region, the latitude effect of isotope value variation is reflected from south to north, but the difference between the south and north indicates that the precipitation amount effect is obvious in the low latitude areas and the temperature effect is obvious in the middle and high latitude areas. In the same latitude area, the inland basins in the northwest arid region have formed high isotope value areas, which indicate that the inland water vapor cycle and secondary evaporation have strong influence on the area. In the Qinghai–Tibet Plateau, the isotope values are generally low due to the influence of high topography (altitude effect), the southwest monsoon weakens from south to north, and the westerly and inland water vapor cycles strengthen. The corresponding isotope values decrease from south to north and then increase. The spatial distribution of precipitation δ18O values can track water vapor from different sources in China.