Divergence of stable isotopes in tap water across China

Stable isotopes in water (e.g., δ2H and δ18O) are important indicators of hydrological and ecological patterns and processes. Tap water can reflect integrated features of regional hydrological processes and human activities. China is a large country with significant meteorological and geographical variations. This report presents the first national-scale survey of Stable Isotopes in Tap Water (SITW) across China. 780 tap water samples have been collected from 95 cities across China from December 2014 to December 2015. (1) Results yielded the Tap Water Line in China is δ2H = 7.72 δ18O + 6.57 (r2 = 0.95). (2) SITW spatial distribution presents typical “continental effect”. (3) SITW seasonal variations indicate clearly regional patterns but no trends at the national level. (4) SITW can be correlated in some parts with geographic or meteorological factors. This work presents the first SITW map in China, which sets up a benchmark for further stable isotopes research across China. This is a critical step toward monitoring and investigating water resources in climate-sensitive regions, so the human-hydrological system. These findings could be used in the future to establish water management strategies at a national or regional scale.

in a sample, and R VSMOW the ratio of heavy to light isotope in international isotopic measurement standard Vienna Standard Mean Ocean Water.
To ensure the accuracy of isotope analysis, each vial was analyzed 6 times. The first three results were abandoned to eliminate memory influence of former sample 27 . During one analysis of a batch of sample vials, the first and last four vials constitutes the standard (Vienna Standard Mean Ocean Water). Regression analysis was conducted to check whether the samples in measure process were problematic 28 . As expected, no samples were identified as problematic.
In order to examine the relationships between tap water isotope and meteorological factors, meteorological data -including the precipitation amount (P, mm), temperature (T, °C), relative humidity (RH, %) and air pressure (PR, kpa) -were collected at observation station in the same city of each sampling location. All the meteorological data were collected from the China Meteorological Data System (http://data.cma.cn/).

Results and Discussion
Spatial pattern of tap water isotopes. There was a large range in δ 18 29 ranged from − 5.86‰ to 20.6‰ with an average of 9.3‰. The Tap Water Line (TWL) of China based on the 780 tap water analyses was: δ 2 H = 7.72δ 18 O + 6.57 (r 2 = 0.95) (Fig. 2). The tap water data clustered near Global Meteoric Water Line (GMWL: δ 2 H = 8δ 18 O + 10) 30 . Both slope and interception in the equation were lower than those in GMWL, which may reflect the effects of evaporation in tap water sources 31 . Compared with Chinese Precipitation Meteoric Water Line 23 , δ 2 H = 7.48δ 18 O + 1.01, TWL exhibited different intercept at 6.57. Although both tap and precipitation datasets were collected across China, the dataset we presented was collected in sequential months from 2014 to 2015. The precipitation data presented in previous study was collected in 29 stations from 2005 to 2010 (no data from 2008). The linear relationship of δ 2 H and δ 18 O in the previous study in the USA 25 collected from 349 tap water samples is: δ 2 H August = 8.02δ 18 O August + 8.21, δ 2 H February = 8.12 δ 18 O February + 9.49. Compared with GWML, the slope of their dataset is quite similar while the interception is a bit lower. Obviously, there is significant difference in tap water isotopic composition between China and USA as a result of different water supply sources.
Spatial patterns in the isotope values were analyzed using Moran's test 32 . Moran's I for δ 2 H and δ 18 O were 0.3 and 0.4, Z = 8.08 and 7.1 respectively, p < 0.01 for both, which means the spatial distribution of tap water isotopes is not random. Figure 3 shows a geospatial interpolation mapping of mean annual δ 18 O, δ 2 H and d-excess in contiguous China. Individual tap water's annual average values are presented on a background colored using Inverse Distance Weighted interpolation model (IDW) in ArcGIS 9.3 (https://www.arcgis.com/features/index. html). In general, tap water isotope values decrease from coastal regions with low latitude and low elevation to inland regions with high latitude and high elevation. This spatial pattern, decrease of isotope values from coastal to inland areas ("continental effect" 33 ) is analogous to results in the previous study in the USA 18 .
The highest δ 18 O and δ 2 H values in annual average (− 4.75‰ and − 30.69‰) appeared in Shanghai on Yangtze River Delta. Other samples with relatively high values were mainly obtained from coastal area in southeastern The elevation map here is presented to give an overview of the China landscape and surrounding environment of the sampling locations. (All of the items were generated with Arcgis 9.3, https://www.arcgis.com/features/index.html). 10 Fig. 1. The extremely low isotope values occurring in these locations could be due to several factors. First, high altitude can lead to extremely low isotopes in precipitation as there is a strong negative correlation between them 34 . Tap water derived from local source that was initially contributed by local precipitation will probably display similar isotope composition at very low ratios. This may, to some extent, explain the extremely low isotope ratios of tap water in Lhasa and Nyingchi (3657 m and 3300 m). Second, in regions with high latitude, e.g., Harbin and Heihe (44.1°N and 50.2°N), isotope ratios in precipitation is strongly linked to local temperatures in high latitudes 35,36 . Tap water derived from regions with high latitude and low temperature tends to have lower isotope values. In both regions mentioned above, high latitude and altitude are related to low temperature which can influence isotope fractionation in precipitation.
In contrast with δ 18 O and δ 2 H, deuterium excess in China shows no clear pattern with extreme high values (> 14‰) found in northwestern arid region (including Xinjiang, Gansu, Qinghai provinces). This is the same

Table 1. Listing of geographical and meteorological information of each sampling locations, including latitude, longitude and elevation above sea level in meters, mean annual temperature (MAT), precipitation (MAP), relative humidity (MARH) and air pressure (MAPR).
finding for extreme low values (< 1‰) found in North China Plain and Inner Mongolia, except for one specific city named Xichang (− 0.77‰) located on Yunnan-Guizhou Plateau in southwestern China (see the dark brown color site in Fig. 1). Standard deviation of monthly isotope values of each site were calculated in order to analyze intra-annual variability of tap water in China. The standard deviation values range from 0.06‰ to 1.79‰, 0.12‰ to 11.83‰ and 0.01‰ to 6.46‰ for δ 18 O, δ 2 H and d-excess, respectively (Table 2). In general, intra-annual variability shows no clear spatial pattern. For certain areas, isotope values exhibit low intra-annual variability, such as Inner Mongolia, Gansu and Qinghai provinces. Sample locations with relatively high intra-annual variability mainly occurred in coastal regions. Similar to δ 2 H and δ 18 O, intra-annual variability of deuterium excess exhibits no clear spatial pattern. Extreme standard deviation value (6.46‰) occurred in Xichang. Moreover, sampling locations with relatively high intra-annual variability centered in western part of the country ranging from 2.5‰ to 3.5‰.

Temporal variability of tap water isotopes. Temporal variability of isotopes in individual tap water
sampling locations was evaluated based on monthly dataset. However, due to certain unavoidable factors including human factors and express delivery's delay in sending and receiving sampling bottles, interval of tap sample acquisition wasn't exactly 30 days but varied from 20 to 40 days. Sampling data series weren't sequential at monthly scale. Therefore, temporal variability was calculated by on-site seasonal comparison: spring (average of March, April and May in 2015) minus winter (average of December in 2014, January and February in 2015), summer (average of June, July and August in 2015) minus spring, autumn (average of September, October and November in 2015) minus summer (see data statistics in Table 3).
Seasonal differences of δ 2 H isotope values spanned 48.51‰ (− 25.99‰ to 22.52‰) with an average of 0.38‰ and a standard deviation of 5.29‰. Seasonal differences of δ 18 O isotope value spanned 5.88‰ (− 2.99‰ to 2.89‰), with an average of 0.02‰ and a standard deviation of 0.78‰. At national scale, there seems no specific pattern of seasonal variability. However, detailed interpretations of seasonal patterns can be found at the regional scale, which is consistent with the findings in precipitation isotope across China by Chen et al. 37 (Fig. 4(a)). Considering the relationship between δ 2 H and δ 18 O, only the δ 2 H plots are shown.
In southeastern regions (Guangxi, Guangdong, Jiangsu, Zhejiang, Shanghai, Fujian, Anhui, Jiangxi, Hunan, Hubei) with a total number of 27 samples locations, most sample locations experienced isotope values rose from winter to spring and dropped from spring to autumn. In general, the maximum isotope values of southeastern region usually occurred in spring and the minimum values occurred in summer or winter.
In northeastern China (Heilongjiang province, Jilin province, Liaoning province and northeast of Inner Mongolia) with 8 samples locations, isotope values in all samples locations except Dalian and Dandong (see rose quartz and apple green color site in Fig. 1) reached the lowest point in late spring or early summer (May or June) and increased to top in late autumn or early winter (November or December) with a spanning range of 6.95‰ in average. Different from the first-sight-guess that extreme isotope values should occur in summer or winter with difference value spanning a large range, for example, stable isotopes in precipitation present regular temporal trends driven by monsoon 37 . Seasonal variability of isotopes in tap water exhibits various pattern with extreme values occurring in various seasons. The reasons might be: a) tap water has mixing water sources as compared to precipitation; and b) there is a lag time between tap water and precipitation. Although only 6 locations on Tibet Plateau provided tap water samples, seasonal trend of isotopes in 5 locations except Nyingchi (see ginger pink color site in Fig. 1) exhibited similar pattern with isotope values decreasing from winter to spring and increasing from summer to autumn. Many factors could contribute to this trend including geographical, climatic, and hydrological factors. Compared to warm regions, the hydrological factors influencing SITW in Tibet Plateau are more complex due to its unique and comprehensive processes happening in cold area, e.g., snow and glacier melting [38][39][40][41] .
These results mean intra-annual variabilities of isotope ratios in tap water are relatively large and the temporal patterns of different regions divided according to the spatial pattern are significantly different. In other words, the temporal patterns of isotopic compositions are, to some extent, correlated with spatial pattern.
Seasonal differences of deuterium excess value spanned 15.61‰ (− 8.12‰ to 7.49‰) with an average of 2.44‰ and a standard deviation of 2.36‰. Deuterium excess is known as providing information about climate conditions of water moisture 42 . Seasonal variability of d-excess is presented in Fig. 4(b). On national scale, deuterium excess values in 76% of the locations increased all the way from winter to summer for about 2.03‰ in average and dropped from summer to autumn for about 1.69‰ in average. Special sample locations with different variation patterns included Heihe in northeast, Korla and Karamay in northwest, 11 locations in north China, Lhasa and Nyingchi on Tibet Plateau and 9 locations in southwest (see color site in Fig. 1). Tap water grabbed from winter or autumn exhibited the most extreme negative d-excess values and lay furthest from GMWL, suggesting a strong evaporative isotopic fractionation of the source waters. While tap samples from summer obtaining the highest d-excess values suggested more evaporated moisture has been added to the atmosphere 43 .
Correlations between isotope values in tap water and environmental variables. Given that isotopes in tap water present various spatial and temporal patterns across China, more detailed work was conducted  to analyze environmental factors influencing tap water isotopes. As demonstrated in many previous studies, isotopes in precipitation 23,44,45 or river 16,46 are strongly correlated to geographical factors (e.g. longitude, latitude, elevation) and climatic factors (e.g., air temperature, precipitation, relative humidity and air pressure et al.) However, tap water does not directly get involved in natural water circulation processes like precipitation, surface water or groundwater. It is a mixture of locally available waters (including rivers, lakes, wells and springs). Therefore, interpretation of tap water isotopes and environmental variables may not be similar to precipitation, which presents 'temperature effect' resulting from different processes of isotopic fractionation 29 . Figure 5 illustrates correlations between mean annual values of δ 18 O and mean annual values of climatic variables or geographic parameters. Note that the elevation data used here is taken from station observation provided by China Meteorological Data System (Table 1). Even though spatial pattern ("continental effect") of isotopes can appear in tap water, the coefficient of determination between δ 18 O and longitude, latitude and elevation were low (r 2 = 0.15, 0.17 and 0.3 for longitude, latitude and elevation, respectively, p < 0.001 for all cases) (Fig. 5(a-c)). Nonetheless, the slope of regression line that reflects elevation effect is − 0.15‰/100 m, which compared well with results of China precipitation δ 18 O values demonstrated by Liu et al. 23 (− 0.13‰/100 m for height).
Correlations between isotopic composition and meteorological factors have been analyzed with 4 extreme low locations (Lhasa, Nyingchi, Heihe and Harbin mentioned in section 3.1) left out (Fig. 5(d-g)). tap water δ 18 O across China had a relatively strong positive correlation with mean annual precipitation (MAP), mean annual   18 2 Similarly, multiple regression model for tap water δ 2 H is as follows: Correlations between mean annual d-excess and environmental variables were also analyzed. However, there are no significant correlations between d-excess and those seven environmental factors with all correlation coefficients lower than 0.1. D-excess in air masses (and hence precipitation) depends on the relative humidity of the air masses at their oceanic origin, the ocean surface temperature, and kinetic isotope effects during evaporation 47 . Given this, it is expected that correlations between d-excess and other environmental factors are weak. In addition, "mixing effect", involving different natural water sources, can also smear the signature leading to such results.
The limitations of this work arise from the data constraints, and the complexity of the natural water cycle and tap water system. First, tap water isotope data requires improvements in sampling duration and spatial coverage to better represent the spatial and temporal pattern across the whole country. This is especially true for the seasonal variability analysis and multi-year observations are preferable. Therefore, current analyses on temporal variability at the seasonal scale might need further refinement. Second, because of the difficulty in sampling concurrent precipitation, surface water, groundwater and examining the complex tap water processing system, we can hardly trace the initial origin of tap water and thus decouple all the mixing signature based on the current data. In this regard, correlations with environmental factors may be informative, but not ideal to investigate the controlling factors of tap water stable isotopic compositions. Further work is needed to better understand the impact of human activity on drinking water system.

Conclusion
To our best knowledge, this study is the first to report tap water isotopic composition over China, which was achieved by establishing a nation-wide volunteer network. Result demonstrated that SITW spatial pattern presents "continental effect" with a decreasing trend in isotopic compositions from coastal regions with low latitude and elevation to inland regions with high latitude and elevation. SITW seasonal trend indicates clearly regional patterns but no trends at the national level, which is consistent with spatial pattern. Also, there are positive correlations between mean annual isotope values and meteorological parameters including precipitation, temperature,