Kinetic processes and stable isotopes in cave dripwaters as indicators of winter severity

We examine how the stable isotope composition of meteoric water is transmitted through soil and epikarst to dripwaters in a cave in western Romania. δ2H and δ18O in precipitation at this site are influenced by temperature and moisture sources (Atlantic and Mediterranean), with lower δ18O in winter and higher in summer. The stable isotope composition of cave dripwaters mimics this seasonal pattern of low and high δ18O, but the onset and end of freezing conditions in the winter season are marked by sharp transitions in the isotopic signature of cave dripwaters of approximately 1 ‰. We interpret these shifts as the result of kinetic isotopic fractionation during the transition phase from water to ice at the onset of freezing conditions and the input of meltwater to the cave at the beginning of the spring season. This process is captured in dripwaters and therefore speleothems from Urșilor Cave, which grew under such dripping points, may have the potential to record past changes in the severity of winters. Similar isotopic changes in dripwaters driven by freeze–thaw processes can affect other caves in areas with winter snow cover, and cave monitoring during such changes is essential in linking the isotopic variability in dripwaters and speleothems to surface climate.

Since the initial recognition of speleothems as palaeoclimate archives (Hendy & Wilson, 1968), oxygen isotopes in speleothems have increasingly been used to reconstruct changes in past climates (McDermott, 2004), providing arguably some of the highest resolution and best dated records of past climate changes (Henderson, 2006).
Most of these reconstructions rely on the relationship between δ 18 O in rainwater and local climate and the transfer of this signal into the cave environment and speleothem carbonate (Fairchild & Baker, 2012;Lachniet, 2009). Insofar as the transfer of this climate signal through the vadose zone, it is necessary to monitor the changes in the stable isotopic signature of both rain and cave waters through time, together with changes in dripwater hydrology (Harmon, 1979).
Studies investigating these variables have focused on the hydrological link between dripwaters and precipitation events. They report either significant seasonal variability of δ 2 H and δ 18 O and a fast response to outside climate (Breitenbach et al., 2015) or significant mixing in the vadose zone with long transfer times between the surface and the cave site, and a corresponding subdued response of water isotopes in dripwaters to high-frequency changes in climate variables (Riechelmann et al., 2011). Knowledge of these response times can then be used to address specific questions related to either short-or long-term climate changes that are likely to be recorded in speleothems.
Robust interpretation of speleothem δ 18 O requires therefore an understanding of both regional influences on rainwater isotopic signature (e.g., source and temperature) and also of the local karst processes (e.g., prior calcite precipitation, kinetic vs. equilibrium fractionation) that can modify this regional signal. The Global Network of Isotopes in Precipitation (GNIP) is a valuable resource used to investigate temporal and spatial variability of stable isotopes in precipitation, but the distribution of GNIP stations in Eastern Europe is very sparse. As a result, the interpretation of stable isotope-based climate proxies in this region is dependent on extrapolation of climate δ 18 O (or δ 2 H) relationships identified in nearby regions. However, as Eastern Europe lies at the intersection of Atlantic, Mediterranean, and continental climatic influences, such extrapolations do not accurately reflect the local isotopic variability in precipitation. Consequently, climate reconstructions using δ 18 O-climate relationships in various sedimentary archives (e.g., speleothems) have difficulties distinguishing between changes in local climate and/or shifts in climatic influences that could be interpreted as changes in climate variables. Romania has a very diverse karst landscape, offering a great potential for reconstructing past climate based on cave deposits (Constantin, Bojar, Lauritzen, & Lundberg, 2007;Drăguşin et al., 2014;Onac et al., 2015;Onac, Constantin, Lundberg, & Lauritzen, 2002;Tămaş, Onac, & Bojar, 2005). However, there are only a few studies available which documented the rainwater δ 18 O variability and corresponding cave dripwater changes in the region (Drăguşin et al., 2017;Perşoiu, Onac, Wynn, Bojar, & Holmgren, 2011), making the climatic interpretation of speleothem δ 18 O from this region ambiguous.
Here, we present an 11-month monitoring study that examines how the isotopic signal of meteoric water is transmitted through soil and epikarst to a cave in western Romania. We use δ 2 H and δ 18 O from precipitation and cave dripping water in the context of karst systems and focus on the processes by which climate signals in stable isotopes are transferred from the atmosphere through the epikarst into cave environments (Mattey et al., 2008;McDermott, Schwarcz, & Rowe, 2005). This study addresses the following goals: firstly, on a global scale, the vast majority of climate proxies are registering either summer climatic conditions (mainly biological proxies) or annual ones, with limited information existing on past winter climate changes (Perșoiu et al., 2017). Knowledge of these would greatly improve our understanding of past climate changes, as well as the ability to forecast future ones, and therefore we focus on isotopic changes in cave dripwaters that occur during the winter season. Secondly, we evaluate the degree to which the kinetic effects associated with water freezing can influence the isotopic signature of speleothems and their climatic interpretation.

| STUDY AREA
Urșilor (Bears) Cave (46°55′ N; 22°56′ E) is located in the Bihor Mountains, western Romania, at 482 m above sea level ( Figure 1). The cave was discovered in 1975 after blasting in a quarry mining recrystallised Upper Jurassic limestone (Rusu, 1981). The total length of the passages, extending over two levels, is 1,500 m. Before discovery, the cave had no natural entrance, but the sizable number of cave bear remains suggest these large mammals had access into the cave during the Last Glacial period. The entrance was subsequently sealed by breakdowns and secondary calcite deposits. The climate in the area is continental temperate, with an annual mean temperature of 9°C. The mean annual precipitation amount is 650 mm, with the highest values occurring in May and June and the lowest in October. Urșilor Cave is one of the main touristic caves in Romania, and it is therefore difficult to establish its natural climate conditions before the cave was developed for tourism. However, monitoring of air temperatures in the cave on three occasions (Racoviţă, Moldovan & Rajka, 1998-1999Racoviţă, Onac, Feier & Menichetti, 2002-2003 and this study found changes in temperature with a maximum amplitude of~2°C between colder (10°C) and warmer months (12°C), although during some milder winters the amplitude of temperature change can be as small as 0.3°C in some parts of the cave. The cave air is characterized by high relative humidity (over 95 %); thus, evaporation effects in the cave are considered minimal (Racoviţă, Onac, Feier & Menichetti, 2002-2003.
Limestone cover above the cave increases progressively from 10 m near its entrance to around 200 m atop of the monitoring site.

| METHODS
Samples of precipitation and cave dripwater were continuously collected between July 2010 and June 2011 and stored at 4°C until stable isotope analysis. Near the cave, a HOBO® multichannel weather station was used to record temperature (0.02°C resolution at 25°C), relative humidity (0.1 % at 25°C or ±2.5 % from 10 % to 90 %), and precipitation amount (rain and snow; ±0.2 mm) every hour.
Precipitation samples were collected every week in 50 ml Nalgene bottles. Cave dripwater was collected using an automated 6712 Portable Teledyne sampler every 100 hr by feeding each bottle with a maximum of 900 ml of dripwater. Any excess water was automatically discarded via an overflow tube. Gemini TinyTag 2 Plus (TGP-4500) data loggers recorded temperature (±0.01°C) throughout the time interval of this research. To prevent evaporation of water samples collected at surface, we used a dip-in-sampler, which was buried in the soil. The device used to collect cave dripwater had dip-in-samplers (with their tubes passing through the bottle cap) locked in a waterproof container. To determine the origin of the air masses delivering precipitation at our site, we used the hybrid single particle Lagrangian integrated trajectory model (HYSPLIT; Stein et al., 2015) with the gridded Global Data Assimilation System meteorological dataset (0.5°resolution).
HYSPLIT has been previously used as a tool for tracking atmospheric circulation and relation with stable isotopes in precipitation (Ersek, Mix, & Clark, 2010;Sjostrom & Welker, 2009). Back trajectories were started at 500 m above ground level and run for a period of 96 hours before arrival. For each precipitation event, we chose the start time of the trajectories to correspond with the highest hourly precipitation amount for that day. A sensitivity test with trajectories started 2 hours before or after the selected time indicated no significant differences in trajectory parameters.

| Precipitation Trajectories
The HYSPLIT back trajectories indicate a predominantly North Atlantic origin for airmasses reaching Urșilor Cave, as expected from

| Stable isotopes in precipitation
The seasonality of δ 2 H and δ 18 O in precipitation is reflective of the temperature influence of the stable isotope composition of rainwater (Figures 2a and 3)

| Effect of freezing on stable isotopes in dripwaters
The stable isotope composition of the dripwater shows a similar trend to that in precipitation, with two intervals of higher values in summer-  (Souchez, Petit, Tison, Jouzel, & Verbeke, 2000). As a result of this process, the heavy ( 18 O and 2 H) isotopes in the water will be preferentially incorporated in the ice (Jouzel & Souchez, 1982;Perşoiu et al., 2011), leaving the remaining water available to reach the cave relatively (to the ice) depleted in these isotopic species. This "freezing signal" is further carried to the cave and recorded by the abrupt drop in dripwater δ 18 O and similarly rapid increase in d-excess. As isotopic fractionation occurs under both equilibrium (at very low freezing rates) and kinetic conditions during the freezing of water (Souchez et al., 2000), the resulting variable fractionation factors led to the alignment of the ice along a line (in a δ 18 O − δ 2 H diagram) with a slope lower than 8 (a so-called "freezing slope"; Jouzel & Souchez, 1982). This, in turn, results in the d-excess of the ice and remaining water having higher values than those in the parent water, as seen in our data: d-excess in precipitation (parent water) has a mean value of 9 ‰ and the associated dripwater a value of about 10 ‰ before freezing, whereas during the interval with negative air temperatures, the d-excess in dripwater (reflecting that of the remaining water after some of the parent one was incorporated in ice at the surface)   (O'Neil, 1968) and in the field (Árnason, 1969;Jouzel & Souchez, 1982;Perşoiu et al., 2011;Souchez & Jouzel, 1984).
Continuous dripping in winter (and partial freezing of precipitation water at the surface) empties the storage zone above the cave. The preferential incorporation of heavy isotopes of H and O in the ice at the surface renders the remaining water progressively depleted in heavy isotopes. The onset of melting in spring sends a pulse of 2 H (and 18 O)-enriched water (Figure 3) that mixes with this stored one and then is slowly released in cave as dripwater. Higher and less variable d-excess values in dripwaters compared with those in precipitation are also indicating that water resulting from ice melting is the main contributor to the recharge of the dripping site.
Freezing at the surface results in low amounts of water available for infiltration and continuous dripping in the cave shows that the cap rock slowly but constantly dries out, as such leaving little water in the matrix available for mixing once surface melting leads to renewed infiltration. Thus, according to a simple mass and isotope balance, this limited amount of water, depleted in heavy isotopes has a reduced influence of the stable isotope composition of new water. Further, this 18 O (and 2 H)-enriched water released after melting, fills the voids in the rock above the drip site, and it slowly released as dripwater. The processes continue until the system is "purged" by the major rain event of 2011.

| Implication for paleoclimatic interpretation of δ 18 O in speleothems
The correlation of local temperature with δ 18 O in precipitation has a slope  Kim and Oneil (1997) is used (−9.54 ‰ and −9.73 ‰ for summer and winter, respectively). We therefore hypothesize