Introduction

Stable hydrogen and oxygen isotopes (δ2H, δ18O and δ17O) can be used as powerful tracers to investigate hydrological processes across multiple spatial scales, including the ecohydrological (e.g., identification of vegetation water sources and partitioning of evapotranspiration) and hydroclimatic processes (e.g., separating hydrographs and quantifying atmospheric processes)1,2,3,4,5,6. The δ2H and δ18O variations of precipitation are mainly determined by temperature in middle and high latitudes7, while precipitation amount is the main determining factor in the tropics8. Variations in the isotopic composition of precipitation are also affected by the source of air masses, elevation of condensation, distance from the coast, and latitude3,7. In addition, for individual precipitation events at a site, the isotopic composition is influenced by synoptic weather patterns, such as the atmospheric conditions at the moisture source, moisture transport trajectories, mixing between vapors from different origins and subcloud processes (e.g., re-evaporation and convection)2,3,9.

During phase change (e.g., evaporation, condensation, and sublimation), two different types of mass-dependent fractionation process may occur between water vapor and condensed water (liquid or ice crystals)7,10. One is equilibrium fractionation, which is driven by the lower saturation vapor pressure of the heavy isotope molecules with respect to the light isotopes11,12. Liquid condensation is generally thought to be a near-equilibrium process that is controlled by local temperature alone11. The other is kinetic fractionation, which is caused by different diffusivities of different isotopes (i.e., isotopically light molecules diffuse faster than those that are isotopically heavier)12,13. Kinetic fractionation, on the other hand, is related to unidirectional and incomplete reactions11 involved in evaporation at the moisture source site, re-evaporation at the precipitation site and solid condensation at supersaturation with respect to ice crystals (e.g., snowflakes and ice formation)7,13,14.

Although individual stable isotope ratios (δ2H and δ18O) of precipitation are informative, a second-order isotopic variable, deuterium excess (d-excess = δ2H − 8 × δ18O)15, can be further utilized to constrain temporal and spatial variations in ecohydrological processes and hydroclimatic conditions15,16,17,18,19. The d-excess is less variable compared with the individual isotopes (δ2H or δ18O) during the equilibrium fractionation because co-variation of δ2H and δ18O is eliminated and it is more sensitive to the kinetic fractionation processes2,20. d-excess therefore provides an additional constraint on conditions at the moisture source and processes that occur along the vapor’s trajectory as it travels from its origin to the precipitation site, including evaporation at the moisture source, condensation in supersaturation conditions, re-evaporation of raindrops, and moisture exchange in the cloud and sub-cloud layer2,7,9,19.

17O, the least abundant stable isotope form of oxygen, holds potential to provide additional constraints on the mechanisms of precipitation formation. Recent developments of high-precision analytical methods (e.g., water fluorination technique) have made it possible to measure changes in 17O despite its low natural abundance. Like d-excess, δ17O and δ18O show different sensitivities to equilibrium and kinetic fractionation processes13, which has led to another second-order parameter, 17O-excess (17O-excess = ln (δ17O + 1) − 0.528 × ln (δ18O + 1))21, as a new hydrological tracer. In theory, the 17O-excess of precipitation is not influenced by moisture source temperature because of similar temperature effects on 17O and 18O22,23,24. Recent studies from Antarctica, however, show that 17O-excess during snow formation under extreme cold condition has a strong sensitivity to atmospheric temperature due to condensation in supersaturation conditions affected by kinetic fractionation10,25,26. Therefore, unlike the d-excess (sensitive to both temperature and relative humidity (RH))27, 17O-excess in precipitation is mainly affected by the RH and insensitive to temperature at the moisture source, though it may be affected by the supersaturation effect under extremely cold conditions (−80 to −15 °C)25,28,29,30. Thus far, the studies of 17O-excess variations in precipitation have mainly focused on high-latitude snow and ice cores10,25,30,31, tropical storms8, and tap water (used as a proxy of precipitation) across the continental United States (U.S.)19. There are a limited number of studies on the meteorological factors that influence precipitation isotope variations in the mid-latitude regions. Therefore, to fill the gap in global precipitation isotope datasets, especially for 17O-excess in the mid-latitude regions, we investigated precipitation isotope dynamics during different seasons and explored rainfall-snowfall variations at one site from the Midwestern U.S. To better understand the formation mechanisms of precipitation and expand the role of 17O-excess as a tracer in investigating various ecohydrological processes at different scales, we examined the relationships between 17O-excess, δ18O and d-excess, and analyzed the relationships between isotope variations and the meteorological factors at the local site.

Materials and Methods

Sampling site

Event-based precipitation samples were collected in Zionsville, Indiana (39.88°N, 86.27°W; 258 m above sea level). Mean annual temperature at the site averaged 10.2 °C, with minimum monthly average temperature in February (−7.2 °C), and maximum average monthly temperature in July (22.2 °C) based on meteorological data from 2014 to 2015 (https://www.wunderground.com). The mean annual precipitation was 953.3 mm, with over 54% of the precipitation occurring between April and July with the highest monthly precipitation occurring in June. Precipitation at the site is influenced by different water vapor sources (Continental, Pacific, Atlantic, Gulf of Mexico, and Arctic)32,33,34, leading to relatively complicated influencing factors of the precipitation formation. Based on the climatology of Indiana, a calendar year was divided into four seasons, with spring as March through May, summer as June through August, fall as September through November, and winter as December through February35.

Precipitation sample collections

In this study, event-based precipitation samples were collected from June 2014 to May 2016. In total we collected 235 precipitation samples consisting of 201 rainfall events and 34 snowfall events. To reduce evaporation effects on isotopes, samples were transferred from the precipitation collector to sealed glass vials (Qorpak Bottles, Fisher Scientific Co. Germany) immediately after each event. The samples were then stored at 4 °C until isotope analysis. If the precipitation event was finished after midnight, sampling was conducted at the earliest possible time in the morning. Snowfall samples were melted in sealed plastic bags, poured into the vials and then stored. Prior to measurements, samples containing impurities were filtered with 0.45 μm syringe filters (Cellulose Nitrate Membrane Filters, GE Healthcare Co. UK) or centrifuged (Iec Centra CL2 Centrifuge, Thermo Electron Co. USA) depending on the size of the impurities.

Isotope analysis

δ2H, δ18O and δ17O measurements

Isotopic ratios (δ2H, δ18O and δ17O) of all the precipitation samples were simultaneously measured using a Triple Water Vapor Isotope Analyzer (T-WVIA-45-EP; Los Gatos Research Inc. (LGR), Mountain View, CA, USA), which is based on Off-Axis Integrated Cavity Output Spectroscopy (OA-ICOS) technique, coupling with a Water Vapor Isotope Standard Source (WVISS, LGR, Mountain View, CA, USA) at IUPUI (Indiana University-Purdue University Indianapolis) Ecohydrology Lab. The specific operational procedure has been described by Tian et al.36, therefore only a brief overview was provided here. In order to achieve higher accuracy and precision, the internal temperature of T-WVIA and WVISS were preheated to 80 °C and 50 °C, respectively, and the Teflon tubing connecting the WVISS and the T-WVIA was heated using pipe-heating cable to avoid condensation of water vapor. According to our previous work36, the higher accuracy and precision of our instruments are generally observed under moderate water vapor concentrations (10000–15000 ppm) for all isotopes, so all the samples were measured under 13000 ppm. Each sample was measured for 2 minutes, and the data output frequency for water isotope measurements was 1 Hz, translating to 120 data points for each sample. To attain more accurate 17O-excess measurements, the 1-Hz data were not averaged over the 2-min interval, the detailed calculation procedure was shown in the “17O-excess data processing” section.

Isotope calibration and normalization

Five commercially available working standards from LGR with known isotopic composition, spanning the entire range of our sample measurements (−154.0‰ to −9.2‰, −19.49‰ to −2.69‰ and −10.30‰ to −1.39‰ for δ2H, δ18O and δ17O, respectively), were analyzed routinely as reference waters after every five precipitation samples to check the instrument performance. In addition, in order to reduce inter-laboratory difference using different technique and calibration methods, all of the isotope ratios were normalized using two international water standards Vienna Standard Mean Ocean Water (VSMOW) and Standard Light Antarctic Precipitation (SLAP) following the procedure described in Schoenemann et al.37:

$${\delta }_{sample/VSMOW-SLAP}^{normalized}={\delta }_{sample/VSMOW}^{measured}\frac{({\delta }_{SLAP/VSMOW}^{assigned})}{({\delta }_{SLAP/VSMOW}^{measured})}$$
(1)

where δ is the δ2H, δ18O or δ17O, and the assigned values of δ2HSLAP/VSMOW, δ18OSLAP/VSMOW and δ17OSLAP/VSMOW are −427.50‰, −55.50‰ and −29.6986‰, respectively. In our study, SLAP2 is used as the replacement water standard for SLAP, and it is not significantly different from SLAP for δ18O or δ17O38. Therefore, SLAP2 is still referred as SLAP hereafter. The two international standards (VSMOW and SLAP) were measured once during each day of the measurements.

17O-excess data processing

Since 17O-excess measurements are two orders of magnitude smaller than traditional δ18O measurements (per meg, i.e., 0.001‰), small peculiarities in either δ18O or δ17O may result in significant 17O-excess error30. To ensure the accuracy of 17O-excess measurements, we used mass-dependent fractionation coefficient (θ = ln (δ17O + 1)/ln (δ18O + 1)), varying slightly depending on the degree fractionation processes, as a quality control filter to check each individual measurement. Based on previous studies, the fractionation coefficient of water was found to be 0.511 ± 0.005 for kinetic transport effects1 and 0.529 ± 0.001 for equilibrium effects39. In addition, it has been shown in previous studies that almost all of the 17O-excess values of global precipitation (e.g., rainfall, snowfall, and ice) fall within the range of −100 to + 100 per meg8,9,19,25,29,31. Therefore, in order to minimize sources of error, any measurements outside the 0.506 and 0.530 range, as well as outside the observed range (−100 to +100 per meg), were removed from the analysis. The final 17O-excess value for every precipitation sample was given as the mean value of quality-controlled data. Using this method, the precision of SLAP was 0.79‰, 0.04‰, 0.02‰ and 3 per meg for δ2H, δ18O, δ17O and 17O-excess, respectively. To check the stability of our instrument precision, we also measured the GISP (Greenland Ice Sheet Precipitation, an international standard) and five commercially available working standards from LGR as mentioned above (−154‰ to −9‰, −20‰ to −3‰ and −10‰ to −1‰ for δ2H, δ18O and δ17O, respectively) on the VSMOW-SLAP scale. The precision of these measurements was better than 0.80‰, 0.06‰, 0.03‰ and 12 per meg for δ2H, δ18O, δ17O and 17O-excess, respectively (Table S1).

δ17O measurements are typically performed using the fluorination method for IRMS technique19,26,27,40, and the water sample are repeatedly measured several times. In addition, when measuring δ17O using the Picarro L2140-i wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) instrument (Picarro Inc., Sunnyvale, CA, USA) at the University of Bern CEP (Climate and Environmental Physics) station, only the last three values were used9. These studies reassure us that data quality control by filtering out measurements could be a suggested procedure for 17O-excess determination. Moreover, the 17O-excess precision of our OA-ICOS technique (2 to 12 per meg) is comparable with IRMS technique (4 to 13 per meg)19,25,29,31,37,41 and CRDS method (< 10 per meg)9,42.

Besides analyzing event-based isotope data, to be comparable with many global precipitation isotope data sets such as Global Network of Isotopes in Precipitation (GNIP), amount-weighted isotopic composition at monthly or seasonal scales was also used in this study, as shown in equation (2).

$${\delta }_{i}=\frac{{\sum }_{i=1}^{n}{\delta }_{i}{P}_{i}}{{\sum }_{i=1}^{n}{P}_{i}}$$
(2)

where \({\delta }_{{i}}\) is the isotopic composition of an individual precipitation event with precipitation amount of \({P}_{i}\), n is the total number of precipitation events in a month or within a season.

Meteorological variables

In order to investigate the effects of meteorological factors on isotopic variations and examine the mechanisms of precipitation formation during different seasons, we used the temperature, RH and precipitation amount at the study site. The meteorological data during the study period were obtained from the Zionsville meteorological station (https://www.wunderground.com). In order to determine whether local meteorological factors affected the isotopic variations in a practical sense, we set the threshold of r being 0.32 (i.e., R2 > 0.10, p < 0.05).

Data availability statement

The datasets generated from the current study are available from the corresponding author on reasonable request.

Results and Discussion

The characteristics of precipitation, temperature and RH

Figure 1 shows the daily and monthly meteorological characteristics (i.e., precipitation, temperature and RH) of the sample collection site between June 2014 and May 2016. All the meteorological variables showed distinct seasonal variations (Table 1). Precipitation amount and frequency were mainly concentrated in the summer (319 mm/35%), less in the spring and fall (276 mm/30% and 189 mm/20%, respectively), and the least in the winter (136 mm/15%). Average summer and winter temperatures were 21.7 °C and −1.6 °C, respectively. Temperatures in the spring (11.2 °C) and fall (11.9 °C) were similar. RH increased from 68.2% in the spring to 72.3% in the summer, then decreased to 69.5% in the fall and then reached a maximum of 77.1% in the winter.

Figure 1
figure 1

Daily (triangle and vertical bars) and monthly (red lines) precipitation, temperature and relatively humidity at Zionsville meteorological station between June 2014 and May 2016.

Table 1 Monthly amount-weighted mean values of δ2H, δ18O, δ17O, 17O-excess and d-excess as well as monthly average temperature, relative humidity and total precipitation in central Indiana (June 2014-May 2016).

Isotopic variations of daily and monthly precipitation at different time scales

Precipitation isotopic variations (δ2H, δ18O and δ17O) and the influencing factors

A wide range of the δ18O values was observed in the daily precipitation data during the study period (−28.10‰ to 3.23‰) (Fig. 2), which is greater than what has been observed in other Midwest regions (e.g., the Chicago area; −18.37‰ to −3.18‰)40. The amount-weighted mean value (−6.54‰) is higher than what observed in Switzerland (−9.05‰) and the continental U.S. (−8.0‰)9,19. Large daily precipitation δ18O variations were also observed within different seasons, showing the largest amplitude in the winter (−28.10‰ to −2.73‰) likely due to the complicated water vapor source. This resembles the results from Bondville, Illinois, which show that winter precipitation can originate from the continental U.S., Gulf of Mexico, Pacific, and Arctic sources32. The smallest amplitude of δ18O variations occurred in the fall (−17.35‰ to −0.88‰), which may reflect relatively stable meteorological factors (Fig. 1) and relatively consistent water vapor sources as observed in Bondville (mainly from Pacific and continental sources)32. The mean values of the δ18O showed a seasonal trend with higher value in the summer (−4.93‰) and lower value in the winter (−10.26‰) (Fig. 2). For the monthly precipitation δ18O trends, the mean value over the study period (−6.61‰) as well as the seasonal trend and values were all similar to the daily precipitation values (Table 1). The δ2H and δ17O variations showed similar trends to δ18O for both daily and monthly scales (Fig. 2 and Table 1).

Figure 2
figure 2

Water stable isotope variations on the event-based sampling (circles, triangles and stars) and corresponding monthly means (red lines) in precipitation between June 2014 and May 2016 in west-central United States. From top to bottom: 17O-excess, d-excess, δ17O, δ18O and δ2H during the individual event.

As demonstrated in many previous studies3,7,9, local meteorological factors play an important role for precipitation isotopic variations in addition to the moisture source influence. Therefore, to better understand what factors affect precipitation isotopic values over the study period and within different seasons, we investigated their relationships with temperature, RH, and precipitation amount at the site of precipitation. The isotopes (δ2H, δ18O and δ17O) of daily and monthly precipitation over the study period were all affected by temperature (r ≈ 0.71 and 0.88 for daily and monthly precipitation, respectively) (Table 2), which is consistent with the traditional “temperature effect” in subtropical and mid-latitude sites (e.g., in Switzerland (r ≈ 0.56 and 0.85))9,15. It is interesting to note that positive correlations between monthly isotopic composition (δ2H, δ18O and δ17O) and precipitation amount (r ≈ 0.53) over the study period were found (Table 2), in contrast to the classic “amount effect” in tropical regions8,15,43,44. In fact, greater precipitation often occurred in the summer due to moisture being sourced from the Gulf of Mexico32,33, the “anti-amount effect” corresponded to the higher temperature in the summer. We therefore suggest that the observed positive relationship between monthly isotopic composition and precipitation amount is the result of the covariance between isotopes and temperature, and isotope variations are mainly affected by temperature during the study period. Moreover, the spring and winter precipitation isotopic compositions exhibited stronger correlations with daily temperature (r ≈ 0.63 and 0.80) than those observed in the summer and fall (r ≈ 0.35 and 0.44) (Table 3), possibly due to the relatively uniform temperature patterns in the summer and fall (Fig. 1)45. The strong correlations and similar sensitivities between monthly isotopic composition and temperature were also observed in the spring, fall and winter (r ≈ 0.85, 0.83 and 0.86) (Table 4). These indicated that aggregation of precipitation isotopic compositions from daily to monthly scale increases the sensitivity to temperature and reduces the difference between seasons. The winter precipitation isotopic compositions were also affected by the daily and monthly RH (r ≈ 0.44 and 0.95) (Tables and 4), which is similar to what reported for a two-year study (monthly scale) in Switzerland (r ≈ 0.40)9.

Table 2 The relationships between the precipitation isotopes (δ2H/δ18O/δ17O), 17O-excess, d-excess and local meteorological parameters (temperature (T), relative humidity (RH) and precipitation amount (P)) at both daily and monthly time scales over the study period.
Table 3 The relationships between the daily precipitation isotopes (δ2H/δ18O/δ17O), 17O-excess, d-excess and local meteorological parameters (temperature (T), relative humidity (RH) and precipitation amount (P)) within different seasons.
Table 4 The relationships between the monthly precipitation isotopes (δ2H/δ18O/δ17O), 17O-excess, d-excess and local meteorological parameters (temperature (T), relative humidity (RH) and precipitation amount (P)) within different seasons.

Variations of 17O-excess (d-excess)

17O-excess (d-excess) values of daily precipitation during the study period (−17 to 64 per meg (–25.79‰ to 24.02‰)) (Fig. 2) are comparable to what is obtained from Switzerland (−26 to 72 per meg (−27.96‰ to 21.95‰))9, while the range is larger than what has been observed across the continental U.S. (tap water, −6 to 43 per meg (−2.5‰ to 17.8‰))19. The mean value of 17O-excess (d-excess) (31 per meg (9.40‰)) is close to the global meteoric waters (35 ± 16 per meg (10‰))19,29, but it is larger than what observed in some mid-latitude regions (e.g., Switzerland (18 per meg)9 and the continental U.S. (17 ± 11 per meg)19), while lower than that reported from Chicago (59 per meg) where the moisture sources of meteoric waters are the Gulf of Mexico and Lake Michigan leading to higher values40. The range of 17O-excess values in the spring (−17 to 64 per meg) and summer (−11 to 62 per meg) were larger than during the winter (8 to 58 per meg), which was the opposite trend for the range of seasonal δ18O variations (Fig. 2). This indicates that 17O-excess brings additional information on precipitation formation, and the moisture source is not the dominant control on winter 17O-excess variations. More intra-seasonal variability is observed in our study compared to the African monsoon region where 17O-excess remains relatively stable before the monsoon onset and slowly changes during the monsoon season8. In addition, mean values of 17O-excess (d-excess) at our study site were the lowest in the spring and summer (30 per meg (8.07‰ and 8.06‰)), whereas values in the fall and winter were higher (36 and 34 per meg (12.18‰ and 11.50‰)). The seasonal pattern of 17O-excess (d-excess) is similar to the trend reported in Switzerland having the lowest value in summer (13 per meg (~7.1‰)) and the highest in winter (25 per meg (~8.1‰)). This demonstrates that 17O-excess variations have obvious seasonal pattern under the relative influence of kinetic and equilibrium fractionations9, corresponding to different slopes of δ′18O-δ′17O (as a proxy of fractionation factor) within different seasons (Fig. 3). d-excess variations indicate less re-evaporation at the precipitation site in the fall and winter, which is also verified by the higher slope and intercept of local meteoric water line (LMWL) between δ2H and δ18O in the fall (7.52/6.93‰) and winter (8.23/13.17‰) (Fig. 4A). For monthly precipitation, 17O-excess values (d-excess) ranged from 21 to 44 per meg (5.55‰ to 17.61‰) during the study period with an average of 31 per meg (9.45‰) (Table 1), similar to the mean value of daily precipitation. The mean value of 17O-excess values in the summer was the lowest (28 per meg), and the value in the winter was the highest (37 per meg), showing the similar seasonal variation trend with daily precipitation, and d-excess as well.

Figure 3
figure 3

The relationships between δ17O and δ18O based on daily (A) and monthly (B) precipitation within different seasons between June 2014 and May 2016.

Figure 4
figure 4

Local meteoric water lines from δ2H and δ18O in daily (A) and monthly (B) precipitation within different seasons between June 2014 and May 2016.

The relationships between 17O-excess, δ18O, and d-excess

Owing to the complexity of the moisture source over the two years, we, for the first time, quantitatively analyzed the relationships between 17O-excess and both δ18O and d-excess of precipitation at different time scales in the mid-latitude region to probe the evaporative conditions at the moisture source. According to the conceptual evaporation model under both steady and non-steady state conditions, if kinetic fractionation associated with evaporation is the sole influencing factor, 17O-excess should be anti-correlated with δ18O, positively correlated with d-excess, and the slope of the latter should be 0.7–2.0 per meg/‰ (or the slopes of δ′18O-δ′17O between 0.5183 and 0.5265), which could also reflect RH at the site of evaporation (i.e., at the moisture source)19.

In our study, the anti-correlation between daily precipitation 17O-excess and δ18O was weak (r = −0.22, p < 0.001) (Fig. 5A) and a weak positive correlation with d-excess was also observed (r = 0.26, p < 0.001, slope = 0.51 ± 0.12 per meg/‰) (Fig. 6A) over the study period. This suggests that multiple factors, in addition to the kinetic fractionation effect associated with evaporation at the oceanic source regions, influence precipitation isotopic compositions1,14,19,29. In addition, daily precipitation 17O-excess was anti-correlated to δ18O in the spring (r = −0.48, p < 0.001) and summer (r = −0.55, p < 0.001) (Fig. 5A). The δ′18O-δ′17O slopes of daily precipitation were 0.5262 (±0.0004) and 0.5254 (±0.0005) for the spring and summer, respectively (Fig. 3A), which are close to the slopes of tap waters (as a proxy of precipitation) from the eastern and western U.S. (0.526–0.527)19. Based on theoretical predictions19, these results suggest that precipitation experienced steady-state evaporation processes with RH between 50% and 85% in the spring and summer leading to the lower 17O-excess. The positive correlation between 17O-excess and d-excess for summer daily precipitation (r = 0.37, p = 0.001, slope = 0.78 ( ± 0.23) per meg/‰) (Fig. 6A) is within the range of the slopes for the theoretical relationships (0.7–2.0 per meg/‰) as mentioned above, further supporting the steady-state kinetic fractionation effect. The slope is similar to what is observed in Africa (0.94–1.04 per meg/‰), where the precipitation experiences steady-state re-evaporation and convective processes8. However, the slope is different from the Gulf states (the most southerly U.S.; 2.5 ± 1.2 per meg/‰), where the precipitation experiences non steady-state re-evaporation processes19. It is interesting that the slope of δ′18O-δ′17O for daily precipitation in the fall (0.5286 ± 0.0005) was close to that in the winter (0.5285 ± 0.0002) (Fig. 3A). Both were similar to the equilibrium fractionation coefficient of the vapor-liquid equilibrium during 10–40 °C (0.529 ± 0.001)39 and the vapor-solid equilibrium between 0 °C and −40 °C (0.5285–0.5290)46. Similar slopes are observed in Chicago precipitation (0.529 ± 0.003) and NEEM (Greenland) snow (0.528 ± 0.001)27,40. In addition, the lower slope in the summer than winter is similar to the result from Switzerland (0.5255 ± 0.0009 and 0.5271 ± 0.0004 for summer and winter, respectively)9, which is due to higher kinetic fractionation processes in the summer. There were no relationships of monthly precipitation between 17O-excess and both δ18O and d-excess within different seasons (p > 0.05) (Figs. 5B and 6B). Moreover, the δ′18O-δ′17O slope of monthly precipitation in the spring (0.5279 (±0.0007) was obviously higher than that of the daily one (Fig. 3B). These demonstrate that only daily precipitation isotopic variations could better reflect evaporation information at the moisture source, and monthly aggregation loses some evaporation information.

Figure 5
figure 5

The relationships between 17O-excess and δ18O based on daily (A) and monthly (B) precipitation within different seasons between June 2014 and May 2016.

Figure 6
figure 6

The relationships between 17O-excess and d-excess based on daily (A) and monthly (B) precipitation within different seasons between June 2014 and May 2016.

Local factors influencing d-excess and 17O-excess

17O-excess variations reflect different precipitation formation mechanisms at the precipitation site for different latitudes, such as raindrop re-evaporation effect (positive correlation with RH) in Africa8, solid condensation under supersaturation (positive correlation with temperature) in the polar regions10,25, and negative correlation with temperature in the mid-latitude9. Therefore, to better identify the influencing factors of d-excess and 17O-excess at the site of precipitation, it is necessary to disentangle their sensitivity to temperature, RH, and precipitation amount over the study period and within different seasons. We found that the d-excess and 17O-excess of daily precipitation over the study period were almost not affected by the local meteorological factors due to the weak correlations (r < = 0.20) (Table 2). However, 41% of variance in d-excess was explained by the monthly precipitation amount (Table 2), similar to what obtained from Switzerland (39%)9. In addition, for d-excess and 17O-excess in precipitation (daily or monthly) over the study period, d-excess has relatively stronger correlations with temperature, RH, and precipitation amount than does 17O-excess (Table 2). Therefore, relative to the d-excess, precipitation 17O-excess over the study period were mainly affected by the atmospheric conditions at the moisture source and along moisture transport trajectories, while almost not affected by the local meteorological factors. This is different from what is observed in Switzerland and it is found 17O-excess contains the information about local monthly temperature9.

Sensitivities of d-excess and 17O-excess to the local meteorological factors within seasons seem to be more complex than over the whole study period. In the summer, d-excess correlated positively with the daily precipitation amount (r = 0.37) (Table 3), similar to what observed in Africa reflecting the “amount effect”, caused by re-evaporation of raindrops at the precipitation site8. In the winter, d-excess was slightly affected by both the daily temperature and RH (r = 0.36/0.37) (Table 3). Only the d-excess in Africa is significantly correlated with RH both at the seasonal scale and during convective processes (r = 0.82/0.90), which indicates re-evaporation is a key controlling process due to the importance of RH in evaporation models8. This means that d-excess in our winter precipitation might be affected by thunderstorms during warm winters (mean temperature was 0.3 °C and the range was −13.3 to 15.6 °C over the two winters) and the associated high RH (83%; 62–96%) (e.g., the thunderstorm occurred on Dec 23th 2015 and the temperature and RH were 11.1 °C and 93%, respectively). In addition, d-excess values in the fall exhibited a strong negative correlation with the monthly precipitation amount (r = −0.92) (Table 4), showing the opposite trend of summer daily precipitation.

It is worth noting that precipitation 17O-excess (daily or monthly) was not affected by local meteorological factors in the spring (p > 0.05) (Tables 3 and 4), therefore, 17O-excess in the spring could be used as a tracer of the evaporative conditions at the moisture source. However, 17O-excess in other seasons was affected by different meteorological factors. For example, 17O-excess in the fall was positively correlated with the local daily temperature (r = 0.35) (Table 3), while a negative correlation was observed in Switzerland based on monthly temperature from 2012–2014 (r = −0.45)9. Only Antarctica shows positive correlations between snow 17O-excess and temperature due to the kinetic fractionation under supersaturation10,25,26. In fact, the fall precipitation is unlikely to be caused by supersaturation due to the higher temperature range (−7 to 27 °C) compared to polar region and abundant condensation nuclei in the mid-latitudes, while higher 17O-excess with high temperature might be due to the continental recycling of the moisture in the fall19. Moreover, a positive relationship was observed between 17O-excess and the daily precipitation amount in the winter (r = 0.32) (Table 3), and the sensitivity was far less than that for monthly precipitation amount in the summer (r = 0.85) (Table 4), demonstrating that the 17O-excess in the winter is less affected by kinetic fractionation than in the summer. In addition, the relationship between 17O-excess in the summer and the monthly precipitation amount is consistent with the trend observed in the Gulf region (r = 0.59, p = 0.05) and in the tropics14,19. The amount effect suggests that the precipitation 17O-excess in the summer may have been affected by stronger kinetic fractionation associated with re-evaporation of raindrops at the precipitation site47. Convective processes should also be considered especially for gentle thunderstorm events, giving rise to lower 17O-excess in the summer (e.g., on July 17, 2015 (−11 per meg) and May 7, 2016 (−17 per meg)), which is similar to what observed in the central U.S.19.

The isotopic characteristics in rainfall and snowfall

As far as we know, there is no previous research on the difference between rainfall and snowfall isotopes in the mid-latitudes. Additionally, previous studies of snow isotopes (e.g., snowfall, snow pits and ice cores) have mainly focused on the polar regions. However, even under similar lower temperature conditions the isotope variations are sensitive to different mechanisms10,25,27. Therefore, to better understand the precipitation mechanisms, it is necessary to study the different forms of precipitation isotopes in the east-central U.S. and compare the snowfall isotopes variations between mid-latitudes and high-latitudes.

The large range (−28.10‰ to −3.07‰) and depleted average value (−13.05‰) of snowfall δ18O, compared with those of rainfall in the site (−18.13‰ to 3.23‰; −6.10‰), may be due to lower temperatures, different air mass trajectories and different moisture source regions28. The slope and intercept of local meteoric water line (LMWL) between δ2H and δ18O for the rainfall (7.33/3.30‰) were both lower than those of the snowfall (7.70/3.58‰) (Fig. 7A), showing stronger re-evaporation effect for rainfall. Most of snow δ18O values in the high-latitudes (e.g., Alert Canada (−39.4‰ to −34.5‰) and Vostok (−60‰ to −50‰)) are lower than ours, possibly due to the polar climate12,40. In addition, δ2H, δ18O and δ17O in our study were sensitive to the daily temperature regardless of rainfall and snowfall (Table 5), and snowfall variations were more sensitive to the temperature (r ≈ 0.73) than the rainfall (r ≈ 0.48), which is also observed between ice cores δ18O and local temperature in Dome A (−60 to −15 °C) and Vostok (−32 to −38 °C), Antarctic13,25.

Figure 7
figure 7

The relationships between δ18O and δ2H (A), δ18O and δ17O (B) in rainfall and snowfall between June 2014 and May 2016.

Table 5 The relationships between the precipitation isotopes (δ2H/δ18O/δ17O), 17O-excess, d-excess and local meteorological parameters (temperature (T), relative humidity (RH) and precipitation amount (P)) for rainfall and snowfall over the study period.

For 17O-excess, the range of values in rainfall (−17 to 64 per meg) was greater than those for snowfall (2 to 54 per meg). The average of snowfall (34 per meg) was slightly higher than that of rainfall (32 per meg), and both are close to the global meteoric waters (35 ± 16 per meg)19,29. The 17O-excess range of rainfall in our study is much larger than what is obtained from African monsoon precipitation (−10 to 20 per meg) experiencing much stronger raindrop re-evaporation8. In addition, the 17O-excess variations of snowfall in this Midwest region seem to be less complex compared with previous studies in polar regions. For example, our 17O-excess average of snowfall is close to what observed in the Greenland snow (35 ± 13 per meg)27. However, the snowfall 17O-excess range varied considerably from 9 to 51 per meg along the East Antarctica traverse, showing a significant decreasing trend along the traverse due to the difference in supersaturation along the air mass trajectories at low temperature25. Although our snowfall 17O-excess range is comparable to them, the snowfall in our site were actually not affected by the supersaturation which would be explained further below.

Similar to the analyses of precipitation at different time scales, we also inferred the fractionation differences at the moisture source between rainfall and snowfall through the conceptual evaporation model. The rainfall 17O-excess exhibited a negative correlation with δ18O (r = −0.39, p < 0.001) and a positive correlation with d-excess (r = 0.35, p < 0.001, slope = 0.68 ( ± 0.13) per meg/‰), while no correlations were observed in snowfall (Fig. 8). This indicates that snowfall is not affected by supersaturation since a positive correlation between 17O-excess and δ18O should appear under supersaturation such as those in Vostok snowfall10. In addition, the rainfall δ′18O-δ′17O slope (0.5265 (±0.0003)) was lower than in snowfall (0.5286 (±0.0004)) (Fig. 7B). The slope of snowfall is close to the equilibrium fractionation coefficient for vapor-solid equilibrium during 0 °C and −40 °C (0.5285–0.5290)46. Our results indicate that the rainfall is affected by the kinetic fractionation during steady-state evaporation processes at the moisture source, while the snowfall seems to be more affected by equilibrium fractionation.

Figure 8
figure 8

The relationships between 17O-excess and both δ18O (A) and d-excess (B) for rainfall and snowfall between June 2014 and May 2016.

The correlation analyses showed that the rainfall 17O-excess and d-excess were not influenced by the local meteorological factors indicated by the weak correlations (|r|< = 0.28) (Table 5), which were slightly different from the precipitation seasonal sensitivities. Compared with rainfall d-excess, 17O-excess were less affected by the local meteorological factors. In addition, there were no correlations between the snowfall 17O-excess and the local meteorological factors (p > 0.05). Therefore, 17O-excess in rainfall and snowfall could be considered as tracers of evaporative conditions at the moisture source in present study.

Conclusions

Ground-based precipitation isotope records in the mid-latitudes, including the U.S. Midwest, are rare and detailed 17O-excess data from the mid-latitude regions is not seen in literature. To fill these knowledge gaps, the isotopic compositions of event-based precipitation including both rainfall and snowfall were monitored at a site in the west-central U.S. The precipitation δ2H, δ18O and δ17O variations were mainly influenced by temperature over the study period. Based on the conceptual evaporation model, the relationships between 17O-excess and both δ18O and d-excess (or δ′18O-δ′17O) indicated that the precipitation in the spring and summer experienced steady-state kinetic fractionation during evaporation at the moisture source, as well as for the rainfall (vs snowfall). The precipitation in the fall and winter, as well as for the snowfall, were mainly affected by the equilibrium fractionation. The precipitation 17O-excess was affected by some local meteorological factors at the seasonal scale (e.g., monthly precipitation amount in the summer) except in the spring. However, 17O-excess of the rainfall and snowfall were not affected by the meteorological factors over the whole study period. Consequently, 17O-excess of rainfall, snowfall and the spring precipitation could be considered as tracers of evaporative conditions at the moisture source in this Midwestern site. The precipitation 17O-excess at different temporal scales provides additional information to better understand the precipitation formation processes in the mid-latitude regions.