Triple oxygen and hydrogen isotopes of gypsum hydration water for 1 quantitative paleo-humidity reconstruction 2 3 Fernando Gázquez 1* , Mario Morellón 2 , Thomas Bauska 1 , Daniel Herwartz 3 , Jakub Surma 3 , Ana 4 Moreno 4 , Michael Staubwasser 3 , Blas Valero-Garcés 4 , Antonio Delgado-Huertas 5 and David A. 5 Hodell 1 6 1 Godwin Laboratory for Palaeoclimate Research. Department of Earth Sciences. University of 7 Cambridge. Downing Street, Cambridge, CB2 3EQ, United Kingdom 8 2 CITIMAC, Facultad de Ciencias. University of Cantabria. Avenida de los Castros s/n, 39005, Santander, 9 Spain 10 3 Institute für Geology und Mineralogy. Universität zu Köln. Greinstrasse. 4-6, 50939, Köln, Germany. 11 4 Department of Environmental Processes and Global Change, Pyrenean Institute of Ecology (IPE) – 12 CSIC, Campus de Aula Dei, Avenida Montañana, 1005, E-50059, Zaragoza, Spain. 13 5 Laboratorio de Biogeoquímica de Isotopos Estables, Instituto Andaluz de Ciencias de la Tierra 14 IACT (CSIC-UGR). Avda. de las Palmeras, 4, 18100, Armilla, Granada, Spain 15 16 *now at School of Earth and Environmental Sciences. University of St. Andrews. St Andrews, KY16 9AL, 17 Scotland, United Kingdom. 18 19 Abstract 20 Atmospheric relative humidity (RH) is an important parameter affecting vegetation yet palaeo-21 proxies for RH are scarce and difficult to calibrate. Here we use triple oxygen ( 17 O and  18 O) 22 and hydrogen (D) isotopes of structurally bounded gypsum hydration water (GHW) extracted 23 from lacustrine gypsum to quantify past changes in paleo-atmospheric RH. An evaporation 24 *Manuscript Click here to view linked References http://ees.elsevier.com/epsl/viewRCResults.aspx?pdf=1&docID=28677&rev=0&fileID=927375&msid={C8898F28-03ED-4411-AA79-7DB638B81743} isotope mass balance model is used together with Monte Carlo simulations to determine the 25 range of climatological conditions that simultaneously satisfy the stable isotope results of GHW, 26 and with statistically robust estimates of uncertainty. We apply this method to reconstruct the 27 isotopic composition of paleo-waters of Lake Estanya (NE Spain) and changes in atmospheric 28 RH over the last glacial termination and Holocene (from ~15 to 0.6 cal. kyrs BP). The isotopic 29 record indicates the driest conditions occurred during the Younger Dryas (YD; ~12-13 cal. kyrs 30 BP). We estimate a RH of ~40-45% during the YD, which is ~30-35% lower than today. 31 Because of the southward displacement of the Polar Front to ~42 o N, it was both windier and 32 drier during the YD than the Bølling-Allerød period and Holocene. Mean atmospheric moisture 33 gradually increased from the Preboreal to Early Holocene (~11 to 8 cal. kyrs BP, 50-60%), 34 reaching 70-75% RH from ~7.5 cal. kyrs BP until present-day. We demonstrate that combining 35 hydrogen and triple oxygen isotopes in GHW provides a powerful tool for quantitative estimates 36 of past changes in relative humidity. 37 38 Keywords: triple oxygen isotopes, gypsum hydration water, relative humidity, lake sediments, 39 Younger Dryas, Late Glacial-Holocene transition. 40 41 1. Introduction 42 The presence of gypsum (CaSO4·2H2O) in lacustrine sediments is commonly interpreted as 43 evidence of dry climatic conditions in the past (Hodell et al., 1995, 2005; 2012; Torfstein et al., 44 2008; Morellón et al., 2009a; Escobar et al., 2012, amongst many others). Evaporation of Ca 2+ - 45 SO4 2- -rich lake waters can lead to gypsum supersaturation under conditions of high evaporation 46 relative to precipitation (inflow). These conditions are generally accompanied by decreased input 47 of fine-grained allochthonous sediments as a result of decreasing runoff, resulting in sediments 48 that are dominantly composed of gypsum. Interbedded layers of gypsum and other “non-49 evaporitic” facies in lakes are often attributed to alternating wet and dry conditions (e.g. Hodell 50 et al., 1995; Ortiz et al., 2006; Morellón et al., 2009a; Escobar et al., 2012; Valero-Garcés et al., 51 2014; Li et al., 2017). 52 The isotopic composition of lake waters is sensitive to long-term changes in the 53 Evaporation/Inflow (E/I) regime and atmospheric relative humidity (RH), among other 54 parameters (Gibson et al., 2016). In addition to E/I and RH, climatic variations recorded in 55 lacustrine carbonates (i.e.  18 O of authigenic carbonates) can be masked by the effect of 56 temperature on the oxygen isotopic value during carbonate precipitation (Hodell et al., 2012). In 57 contrast, structurally bounded gypsum hydration water (GHW) can be used to reconstruct the 58 isotopic value of paleo-lake waters with little to no effect of temperature. The fractionation 59 factors for oxygen and hydrogen isotopes between the free solution and GHW are largely 60 independent of temperature in the range of most lakes (e.g. 10-35 o C; Gázquez et al., 2017a). 61 Thus, the oxygen and hydrogen isotopes ( 18 O and D) of GHW can be used to infer the isotopic 62 composition of paleo-lake waters at the time of gypsum precipitation (Hodell et al., 2012; Grauel 63 et al., 2016; Li et al., 2017). GHW retains the isotopic values of the parent solution provided that 64 it has not been altered by post-depositional processes (e.g. exposure to temperature >50 o C after 65 deposition, solution-reprecipitation, etc.). Whether the original isotopic composition of GHW has 66 been preserved or not must be evaluated on a case-by-case basis (Hodell et al., 2012; Evans et 67 al., 2015; Gázquez et al., 2017a). 68 Recent analytical developments permit precise measurements of triple oxygen isotopes 69 ( 17 O/ 18 O/ 16 O), and the derived parameter 17 O-excess (also called Δ17 O), in natural waters (Luz 70 and Barkan, 2010; Steig et al., 2014) and GHW (Gázquez et al., 2015) with precision better than 71 ±0.01‰ (i.e. 10 per meg; ±1). This parameter is defined as: 72 17 O-excess = ln( 17 O + 1) – 0.528 ln( 18 O + 1) (Eq. 1) 73 where: 74  17 O and  18 O denote the 17 O/ 16 O and 18 O/ 16 O in water standardized to V-SMOW (Barkan and 75 Luz, 2005; Luz and Barkan, 2010; Schoenemann et al., 2013). The value of 0.528 has been 76 proposed to describe the  17 O and  18 O relationship in rainwater worldwide (Luz and Barkan, 77 2010). The 17 O-excess averages ~37 per meg in modern meteoric waters and shows lower values 78 in evaporated water (Barkan and Luz, 2010; Steig et al., 2014; Surma et al., 2015). The trajectory 79 of  18 O and 17 O-excess in evaporated water is relatively insensitive to temperature and salinities 80 below 100 g/l (Barkan and Luz, 2010; Passey et al., 2014); however, it is significantly affected 81 by other parameters such as the hydrological balance of the water body and atmospheric RH 82 (Surma et al., 2015; Gázquez et al., 2017b; Herwartz et al., 2017; see Fig. 1). 83 Despite the potential of lake sediments as palaeoclimatic archives, stable isotopes in inorganic 84 and organic proxies often allow only qualitative interpretation of past hydrological changes. 85 Quantitative reconstructions from isotope proxy data, including changes in atmospheric relative 86 humidity, have been difficult to achieve and calibrate. Here we evaluate the potential of using 87 triple oxygen and hydrogen isotopes in lacustrine GHW to quantify changes in atmospheric 88 relative humidity in the past. We use a Raleigh evaporation isotope mass balance (IMB) to 89 estimate quantitatively climatic conditions at the time of gypsum precipitation. Monte Carlo 90 simulations are used to find the most probable solution to the model and evaluate uncertainties 91 for RH. We apply this method to isotopic data ( 17 O,  18 O and D, and derived d-excess and 92 17 O-excess) of gypsum hydration water from Lake Estanya (Southern Pre-Pyrenees, NE Spain) 93 to infer climate during the Late Glacial and Holocene periods (ca. 15 cal. kyrs BP to 0.6 cal. kyrs 94 BP). We model the isotopic values of paleo-lake Estanya under different 95 environmental/geochemical scenarios. We compare the isotopic results and derived RH values 96 with previous sedimentological and geochemical proxies in the lake sequence (Morellón et al., 97 2009b), as well as other biological indicators such as pollen, diatoms and chironomids (Morellón 98 et al., 2011; Vegas-Vilarrùbia et al., 2013; González-Sampériz, 2017). Lastly, we discuss more 99 generally the potential application of the method to other lakes. 100 2. Approach and model 101 The oxygen ( 17 O and  18 O) and hydrogen (D) isotopic composition of lake waters increase 102 with more arid conditions and higher evaporation relative to inflow (E/I). Each isotope ratio 103 follows a slightly different fractionation leading to variability in d-excess and 17 O-excess 104 parameters (Surma, 2015; Gibson et al., 2016; Herwartz et al., 2017; and references therein). The 105 isotopic evolution of water during evaporation (e.g.  18 O vs 17 O-excess and  18 O vs d-excess) 106 depends on the isotopic composition of the initial water (inflow), temperature, relative humidity, 107 the isotopic composition of the water vapor in equilibrium with the liquid water and the ratio of 108 water loss by evaporation (E) with respect to the inflow (I), with the remainder lost as outflow. 109 The process is described by the expression (Criss, 1999): 110 111 (Eq. 2) 112 113 Where *RWS is the isotopic ratio of the evaporated water. * 0 evap is the effective fractionation 114 factor, calculated as a product of the equilibrium fractionation factor (* 0 eq) and the diffusive 115 fractionation factor (* 0 diff) between the liquid water and vapor. The parameter h is the relative 116 humidity of air (0 to 1). RWI is the isotopic ratio of the input prior to evaporation (i.e., the inflow 117 to the lake). Rv is the isotopic ratio of the vapor and depends on the degree to which the 118 atmospheric water vapor (veq) is in equilibrium with RWI (Gibson et al., 2016), where: 119 Rv = RWI*( 0 eq*veq) (Eq. 3) 120 E/I represents the fraction of water loss by evaporation with respect to the inflow from the 121 system (e.g. E/I =0 means no evaporation whereas E/I =1 means all the water is lost to 122 evaporation; i.e. there is no outflow). This model assumes homogeneous isotopic composition of 123 both the liquid and vapor phases. 124 Equilibrium fractionation factors for  18 O and D are well known and calculated here as a 125 function of temperature using the equations of Horita and Wesolowski (1994).    0 eq is 126 calculated as    0 eq =    0 eq θ , where θ is 0.529 (Barkan and Luz, 2005). Kinetic fractionation 127 during evaporation under natural conditions is not as strongly constrained as equilibrium 128 fractionation. Here we use a combination of natural and laboratory experiments to calculate  0 diff 129 (Landais et al., 2006; Barkan and Luz, 2007, Luz et al., 2009). First,    0 diff varies as a function 130 wind driven turbulence (Dongmann et al., 1974; Uemura et al., 2010; Haese et al., 2013) (see 131 discussion section) and is calculated as: 132    0 diff = 1.0283 w (Eq. 4) 133 where w varies between 0.5 (pure turbulent mixing;    0 diff = 1.0141) and 1.0 (pure diffusion; 134    0 diff = 1.0283). 135 D 0 diff varies as a function of    0 diff and temperature (T, in o C) based on experiments by Luz 136 et al. (2009), where: 137 D 0 diff = (1.25-0.02 T) (   0 diff -1) +1 (Eq. 5) 138 139    0 diff is calculated as    0 diff =    0 diff θ , where θ is 0.5185 (Landais et al., 2006; Barkan 140 and Luz, 2007). 141 In  18 O- 17 O-excess and  18 O-d-excess space (Fig. 1), the predicted trends of waters undergoing 142 evaporation in partial equilibrium with atmospheric vapor take the form of curves. We see that 143 both 17 O-excess and d-excess are largely sensitive to RH during evaporation (Fig. 1A and F), 144 whereas their sensitivities to temperature are relatively small, especially for 17 O-excess (Fig. 1B) 145 (Landais et al., 2006; Passey et al., 2014; Surma et al., 2015; Gázquez et al., 2017b; Herwartz et 146 al., 2017). Both cross-plots are moderately sensitive to turbulences (winds) on the water surface 147 during evaporation (Fig. 1C and H) and to the isotopic composition of the atmospheric water 148 vapor (Fig. 1D and I). Also, both isotopic systems are very sensitive to the proportion of water 149 loss by evaporation with respect to the input (E/I) (Fig. 1E and J). 150 In summary, our isotopic model is based on three equations where  17 O,  18 O and D are known 151 (measured), four variables that can be constrained by modern estimates ( 17 O,  18 O and D of 152 the inflow and lake temperature), two poorly constrained but minor variables (turbulence and 153 vapor-precipitation equilibrium) and two significant unknowns (E/I and h). Estimating E/I and h 154 in the past requires some assumptions about the errors in the unknown variables and an 155 understanding of how these variables co-vary and introduce uncertainty in the model results (see 156 discussion section). 157 158 Monte Carlo simulations performed in Matlab® is used in our method to find the possible model 159 solutions that satisfy the isotopic composition of the modern and paleo-lake water given the 160 uncertainty in both the paleo-environmental parameters considered and the analytical errors. The 161 approach is represented graphically in Figure S3. Briefly, a range of model inputs is selected 162 based on conservative estimates of their distributions in the modern and potential to change in 163 the past. The distribution of model inputs is uniform and thus produces a largely uniform set of 164 model solutions. The error in each model solution is then calculated relative to the mean and 1-165 sigma standard deviation (1SD) of each individual data point. The normalized errors for  18 O, 166 D, and 17 O-excess are then combined to arrive at a total error (if no 17 O-excess data exists it is 167 excluded from the total error). Only those model solutions that fall within the 1SD are then 168 selected. In the three-dimensional space of the  18 O, D, 17 O-excess, this can be visualised as 169 selecting all the data points that fall within an ellipsoid with axes that extend to the 1SD 170 analytical error in each parameter. From the subset of simulations, the mean and range of model 171 inputs (e.g. relative humidity) that are constituent with the lake water isotopes can derived. 172 173 3. Materials and methods 174 Balsas de Estanya is a karstic lake complex located at the foothills of the Southern Pyrenees 175 (42º02’N, 0º32’E) at 670 m a.s.l. It is a relatively small endorheic basin of 2.45 km 2 (Fig. 2) that 176 comprises multiple bodies of water. The largest and deepest lake (Estanque Grande de Abajo) 177 has been studied extensively for paleoclimate and paleolimnological reconstruction (Morellón et 178 al., 2009b; 2011; and references therein) (Fig. 2 and supplementary material). Twenty-nine 179 gypsum samples were collected from the ca. 11-m long composite sequence of Lake Estanya, 180 which is comprised of a combination of cores LEG04-1A-1K and EST06-1A-1U (Morellón et 181 al., 2009b). The cores were recovered from the deepest areas of Estanque Grande de Abajo (Fig. 182 2). The age model is based on radiocarbon, 137 Cs and 210 Pb and lithostratigraphy as previously 183 described by Morellón et al., (2009b) and Vegas-Vilarrùbia et al. (2013) (see supplementary 184 material). 185 GHW was extracted by heating the powdered gypsum in vacuo using a bespoke offline system 186 consisting of six vacuum lines contained within a modified gas chromatography (GC) oven in the 187 Godwin Laboratory at the University of Cambridge (UK) (Gázquez et al., 2015). Oxygen ( 17 O 188 and 18 O) and hydrogen (D) isotopes of the hydration water were measured simultaneously by 189 cavity ringdown spectroscopy (CRDS) using a L2140-i Picarro water isotope analyzer (Gázquez 190 et al., 2015 and supplementary material for details). All results are reported in parts per thousand 191 (‰) relative to V-SMOW. The uncertainty of the method was ±0.05‰ for  17 O, ±0.1‰ for  18 O 192 and ±0.6‰ for D, ±0.8‰ for d-excess and ±8 per meg for 17 O-excess (1SD). 193 Additionally, the hydration water of two samples were also analyzed for  17 O and  18 O using a 194 modified version of the fluorination-IRMS method of Barkan and Luz (2005) at the Institute for 195 Geology and Mineralogy at the University of Cologne, Germany (Surma et al., 2015; Gázquez et 196 al., 2015; Herwartz et al., 2017). Rain (n=59) and lake waters (n=61) collected between 2001 and 197 2012 were analyzed for oxygen and hydrogen stable isotopes (Tables S2, S3, S4 and S5 in 198 supplementary material). 199 4. Results 200 Twenty-nine gypsum samples, ranging in age from 14.7 to 0.6 cal. kyrs BP, were analyzed for 201 stable isotopes in GHW. The  17 O varies from 3.8‰ to 7.1‰,  18 O from 6.3‰ to 14.9‰ and D 202 from −26.1‰ to 3.0‰ (Table S1 and Fig. S2 in supplementary material). The lowest values 203 correspond to gypsum samples at 136 cm below lake floor; ca. 620 cal. yr BP) and the highest 204 values to gypsum at 588 cm depth; ~12 cal. kyrs BP). 205 The oxygen and hydrogen isotope composition of the parent water from which the gypsum 206 formed is calculated from GHW using recently revised fractionation factors (gypsum-water) 207 (Gázquez et al., 2017a) that are more precise and accurate than previous values (Gonfiantini and 208 Fontes, 1963; Fontes and Gonfiantini, 1967; Hodell et al., 2012). We use  18 Ogypsum-water of 209 1.00355 and Dgypsum-water of 0.979 corresponding to a temperature of 15 o C, representing roughly 210 the modern mean temperature of the lake water (~12.5 o C). Note that both,  18 Ogypsum-water and 211 Dgypsum-water are largely unaffected by temperature in the range from 10 o C to 35 o C (Gázquez et 212 al., 2017a). The use of temperatures that are 10 o C higher or lower changes the  18 O values by 213 only ~±0.1‰ and D by ~±2‰, which is not very significant relative to the analytical precision 214 of our method. 215 The relation between  17 Ogypsum-water and  18 Ogypsum-water is given by the parameter θ ( 17 Ogypsum-216 water =  18 Ogypsum-water θ ), which has been found to be 0.5297±0.0012 and does not depend on 217 temperature between 3 and 55 o C (Gázquez et al., 2017a). Therefore, we use  17 Ogypsum-water of 218 1.00188. Using these alpha values, we found that the paleo-lake water (i.e., GHW corrected for 219 fractionation) plot on an evaporation line with slope of 3.4 (Fig. 3). This evaporation line is 220 comprised of paleo-lake waters from different ages that evaporated under different 221 environmental conditions. Therefore, the slope of this line does not convey a unique paleo-222 hydrological significance. 223 From 14.7 to 13.3 cal. kyrs BP during the Bølling-Allerød (B-A) period, the  18 O values of the 224 lake water increased gradually from 7.8‰ at 14.7 cal. kyrs BP to 10.6‰ at 13.3 cal. kyrs BP, 225 while D increased from 12.1‰ to 21.9‰. During the same period, d-excess varied from -226 50.4‰ to -62.6‰. The  18 O and D of the lake water shows the highest values of the entire 227 record at ca. 12 cal. kyrs BP (11.3‰ and 23.7‰, respectively) during the Younger Dryas (YD) 228 Chronozone. This time also marked the lowest d-excess values (-69‰) (Table S1). During the 229 following Preboreal-Holocene period (from 11.7 to 7.5 cal. kyrs BP),  18 O and D values 230 decreased to ~5.5‰ and ~2.4‰, respectively. Finally, the isotopic composition of the lake water 231 reached full Holocene conditions after ~7.5 cal. kyrs BP ( 18 O of 4.3±0.7‰ and D of -232 1.5±2.9‰), showing similar values to modern Lake Estanya water ( 18 O of 3.6±0.7‰ and D of 233 -2.4±7.1‰). The 17 O-excess of paleolake water during the Holocene ranged from -63 to -46 per 234 meg, also resembling modern values (-51 per meg). More negative 17 O-excess values were 235 recorded during the Preboreal-Early Holocene (-103 to -94 per meg), the YD (-82 per meg) and 236 the B-A period (-67 per meg). 237 238 5. Discussion 239 5.1. Reliability of GHW results 240 Recent stable isotope studies of GHW in lakes have produced relevant paleoclimatic records that 241 closely agree with other local and regional climatic proxies (Hodell et al., 2012; Grauel et al., 242 2016; Li et al., 2017 and the present study). Such excellent correlations undoubtedly indicate 243 that, at least in some cases, the primary isotopic composition of GHW is preserved in time. 244 Investigations on Messinian gypsum deposits (ca. 5.9 Ma) also suggest no isotopic exchange or 245 alteration of the primary isotopic signal (Evans et al. 2015). However, it must be considered that 246 in other sedimentary sequence, GHW may have undergone isotopic modification, for example by 247 dehydration/rehydration cycles as a result exposure of gypsum to temperature >50 o C (e.g. burial 248 and exhumation cycles). Therefore, the reliability of stable isotopes in GHW to reconstruct the 249 isotopic composition of paleo-waters should be evaluated on a case-by-case basis. 250 There are several lines of evidence that GHW in Lake Estanya preserves its primary isotopic 251 signature. After applying fractionation factors, the values of the Mid- Late-Holocene paleolake 252 waters roughly match the modern lake water; however, the Early Holocene and Late Glacial 253 paleolake waters (~7.5 to ~15 cal. kyrs BP) show considerably more enriched values. If the 254 GHW had exchanged with sediment pore water, we would expect a relatively homogeneous 255 isotopic profile with values similar to the current lake water. Because the burial depth is shallow 256 and sediments are porous we expect any isotopic gradients in pore water to be strongly 257 attenuated by diffusion and advection with overlying lake water. 258 The isotopic changes in GHW through the late Glacial-Holocene transition and the Holocene 259 strongly correlate with major climatic changes recorded by other regional and local paleoclimate 260 archives, including several sedimentary proxies in Lake Estanya. Accordingly, we suggest the 261 hydration water of gypsum deposits in Lake Estanya reflects the isotopic composition of the 262 paleo-lake water during the latter part of the last deglaciation and Holocene. 263 264 5.2. Determining RH from isotopic analysis of GHW 265 The ability of GHW to record the isotopic composition of the original fluid, with little to no 266 effect of temperature, makes it a near direct proxy for the isotopic composition of paleowater. 267 The method presented here permits  17 O,  18 O and D, and derived d-excess and 17 O-excess to 268 be determined simultaneously in the same sample. The isotopic composition of paleo-lake water 269 can then be used to model the hydrologic parameters of the basin and climatic conditions at the 270 time of gypsum precipitation. Importantly, the ability of this method to reconstruct the 17 O-271 excess of paleo-waters, which is dependent of RH and practically insensitive to temperature 272 during water evaporation, constitutes a powerful tool for paleo-hydrologic reconstructions. 273 When modelling the isotopic composition of paleo-lakes to fit the GHW data, several parameters 274 must be known or assumed. The uncertainty in some variables, including temperature and the 275 isotopic composition of the freshwater member, have relatively little effect on the results of the 276 model (Fig. 4 and 5B). For example, a temperature change of 3-5 o C, as expected for the last 277 Glacial-Holocene transition in some regions (see section 5.3), barely affects the model results for 278 17 O-excess (up to ~±2 per meg in a terminal lake), whereas d-excess changes by up to ~3‰ in a 279 terminal lake, when keeping all other parameters constant. Importantly, the model solution must 280 satisfy both 17 O-excess and d-excess of the same paleo-water. 281 The isotopic composition of rainfall varied between glacial and interglacial periods in most 282 regions, as recorded by speleothems, paleo-groundwaters and ice cores. For example,  18 O of 283 freshwater in the western-Mediterranean region increased up to ~1‰, although practically no 284 change has been observed in south Iberia during the last glacial-Holocene transition (Jasechko et 285 al., 2015) (see section 4.2. in supplementary material). As seen in Fig. 5B, when keeping other 286 parameter constant, a 1‰ change in  18 O of the input will result in uncertainties of 3 to 4.6% in 287 the modeled RH. However, it worth noting that larger change in  18 O of the rainfall may be 288 expected in other regions; for example, in areas affected by monsoonal systems and large 289 variations in the “amount effect” over glacial-interglacial cycles (e.g. southern Asia; Kathayat et 290 al., 2016). Therefore, uncertainties in the isotopic composition of the freshwater member must be 291 considered as a potential source of error for quantitative paleo-humidity estimates when using 292 this method. 293 The isotopic composition of the modern atmospheric vapor and how it varies with time is 294 unknown in most regions, nor there are estimates available for this parameter in the past. 295 However, a recent study suggests that the isotopic composition of atmospheric vapor is often in 296 partial equilibrium with that of local freshwater. A degree of equilibrium of 75% seems 297 reasonable for most tropical and intertropical regions (Gibson et al., 2016). Our IMB model does 298 not reproduce the measured  18 O,  17 O and D ratios for any reasonable set of input parameters 299 when full equilibrium is assumed. The model is relatively sensitive to the isotopic composition 300 of the vapor, as shown in Fig 1D. This must be considered when modeling the isotopic 301 composition of lakes in coastal areas that can be affected by advection of marine air masses, 302 whose isotopic composition is in equilibrium with seawater instead of freshwater. 303 The effect of turbulence (e.g., wind) on the isotopic equilibrium between water and vapor is 304 accounted for in our model by replacing *αdiff by (*αdiff) w , where the exponent ‘w’ is set between 305 0.5 (pure turbulence) and 1 (no wind) (Dongmann et al., 1974; Uemura et al., 2010; Haese et al., 306 2013). The relationship between this parameter and the wind speed is not well constrained; 307 however, it is known that the proportion of *αdiff may be suppressed by turbulent flow induced 308 by wind (e.g. Horita et al. 2008). As exemplified in Fig. 4 (see section 5.3 and supplementary 309 material for details), when turbulence is not considered, the model yields d-excess and 17 O-310 excess values that are systematically too low compared to the analytical data for some periods 311 (i.e. Younger Dryas; ~12 ka). This offset can be corrected by reducing the value of the exponent 312 ‘w’ for periods that are documented to have been windier than the average. 313 The hydrologic balance of the lake (E/I) controls the isotopic composition of the water. Lakes 314 with high E/I values (i.e. lakes of dry regions) may also show high salinities due to accumulation 315 of salts in the basin. However, the salt effect on the IMB becomes significant only at >100 g/l 316 (Sofer and Gat, 1972; Criss 1999; Herwartz et al., 2017). This salinity values may be reached in 317 some hypersaline chlorine-rich lakes, for which a salinity correction would be needed (Herwartz 318 et al., 2017). Nevertheless, gypsum precipitation does not necessary occur in high-salinity 319 environments, but often takes place in freshwater lakes saturated in calcium sulfate with 320 relatively low salinities, often ~3-4 g/l (e.g. Hodell et al., 2005; Perez-Bielsa, 2013). The E/I of 321 the lake has a large impact on the IMB. The E/I in modern lakes can be asserted by a simple 322 mass balance of conservative elements in water (e.g. sodium chloride), but this parameter in the 323 past is generally unknown. Figure 5A shows that when E/I exceeds 0.75, changes in this 324 parameter barely affect the RH values derived from the model. We estimate the errors for the RH 325 to be smaller than 5% (1SD) when the lake approached terminal conditions, as required for 326 saturation in gypsum of water. In contrast, when the model is forced to E/I <0.5 the scatter of RH 327 values increases (ca. ±15%, 1SD), suggesting that the resolution of our method for RH 328 estimation is better for lake systems close to closed conditions (all the water loss by evaporation 329 and practically not outflow) than for throughflow lakes. This is because of the  18 O- 17 O-excess 330 and  18 O -d-excess trajectories of evaporated waters under different RH diverge considerably 331 when E/I approaches 1 (Fig. 1). In contrast, the isotopic trajectory of water in a throughflow lake 332 (e.g. E/I<0.5) barely differs when evaporation occurs under different conditions of RH. This 333 indicates the method described here is especially suitable for lakes in which gypsum formed 334 under arid or semiarid conditions (Surma et al., 2015). 335 In summary, when the model is forced to match both the 17 O-excess and d-excess of the paleo-336 water measured in GHW and model inputs are selected based on conservative estimates and 337 appropriate errors, the derived uncertainty in RH can be as low as ±3% (1). The model is 338 insensitive to temperature changes, whereas uncertainties in the isotopic composition of the 339 rainfall can have a significant effect, especially in regions where the isotopic composition of 340 rainfall is highly variable. The accuracy of the method is best when applied to lakes under 341 arid/semiarid climate (RH<70%) and elevated E/I (hydrologically closed basins). Most of these 342 conditions are met for Lake Estanya where we have applied the method to estimate RH changes 343 during the last glacial termination and Holocene. 344 345 5.3. Application to Lake Estanya 346 We applied the approach described above to reconstruct paleoclimate at Lake Estanya in the 347 Southern Pyrenees during the late Glacial-Holocene transition and the Holocene. We interpret 348 past changes in the isotopic composition in GHW of Lake Estanya in terms of changing RH (see 349 supplementary material for detailed rationale about the environmental parameters selected for the 350 model). 351 352 5.3.1. Bølling-Allerød period 353 Between ~15 and ~13 cal. kyrs BP, coinciding with the Bølling-Allerød (B-A) period, the lake 354 showed intermediate  18 O and D values compared with the later stages. This indicates a more 355 positive water balance compared to the subsequent period (i.e. 12.8-11.6 cal. kyrs BP; Younger 356 Dryas). Our model based on 17 O-excess and d-excess suggests that RH during the B-A period 357 was ~55-65%. This is ~10-15% less than modern conditions in the Estanya region (~70-75%). 358 This finding is consistent with comparatively lower water salinity and greater productivity in the 359 paleo-lake than during the YD, inferred from the elemental composition (XRF) of the sediments 360 and  13 C of organic matter, respectively (Morellón et al., 2009b) (Fig. 6). During the B-A period, 361 a trend towards heavier  18 O and D values was recorded, reaching a relative maximum at ca. 362 13.2 cal. kyrs BP, coinciding with a cold period. 363 364 365 5.3.2. Younger Dryas 366 Relatively enriched  18 O and D values (11.3‰ and 23.7‰, respectively) and lower d-excess (-367 69‰) in the lake water recorded during the YD are also in good agreement with higher E/I 368 compared to previous and later stages. Accordingly, a maximum in water salinity during the YD 369 is also evidenced by the greatest contents of S and Ca 2- in this section of the core (XRF data in 370 Morellón et al., 2009b). The modelled 17 O-excess and d-excess of the paleolake water during the 371 YD indicate that atmospheric RH decreased to 40-45% during the aridity peak at ~12 cal. kyrs 372 BP. This is ~10-15% less than during the previous B-A period and ~30-35% less compared to 373 present. These values are also consistent with recent studies of biomarkes in Lake Meerfelder 374 Maar (Germany) that suggest RH decreased by 8-15% during YD compared with the previous 375 B/A period (Rach et al. 2017). Also, these results are in good agreement with previous studies 376 suggesting the YD in NE Spain was characterized by cold and arid conditions, with particularly 377 extreme conditions in higher altitudes of the Pyrenees (González-Sampériz et al., 2006). 378 Accordingly, generally drier conditions were recorded elsewhere on the Iberian Peninsula 379 (Moreno et al., 2012). A high-resolution speleothem record from El Seso Cave (Southern 380 Pyrenees) reveals a modest cooling of 1.3ºC compared with other circum-Iberian sea surface 381 temperature reconstructions (Cacho et al., 1999; Eynaud et al., 2009), and a significant decrease 382 in rainfall during the first part of the YD (12.9-12.5 cal krys BP) followed by a progressive 383 increase in humidity afterwards. 384 Drier conditions throughout the YD have also been inferred from other paleoclimatic sequences 385 in Iberia (Moreno et al., 2012, Garcia-Ruiz et al., 2016; González-Sampériz et al., 2017). The 386 maximum southward migration of the polar front during the YD reached 42ºN (Broecker et al., 387 1988; Lane et al., 2013), approximately the latitude of the Southern Pyrenees and Lake Estanya. 388 Marine records from the Iberian Margin found a pronounced cooling during this period, even 389 more intense than in the LGM (Eynaud et al., 2009 and references therein). Furthermore, the 390 existence of loess deposits (13-10 cal. kyrs BP) in Central Spain (Bateman and Díez-Herrero, 391 2001) also supports an increase in aridity and perhaps wind speed in Western Europe during the 392 YD (Brauer et al., 2008), as suggested by our model results. 393 394 5.3.3. Holocene 395 The Early Holocene period (11.7 to 7.5 cal. kyrs BP) was characterized by a decrease in  18 O 396 and D values (by ~9‰ and ~20‰, respectively). This indicates a more positive water balance 397 (lower E/I) compared to the YD. Our model indicates that RH increased to 50-60% during the 398 Early Holocene (~7.5-11 cal. kyrs BP). Pollen-based vegetation reconstructions indicate 399 relatively dry conditions during the transition to the Holocene marked by increasing Juniper sp. 400 and decreasing mesophytes (González-Sampériz et al., 2017). The isotope values reveal a 401 comparatively large increase in humidity relative to the YD, which is not reconstructed by 402 sedimentary facies and palynology but is in agreement with other paleohydrological records of 403 NE Spain and other regions from the Iberian Peninsula (Moreno et al., 2012; Morellón et al, 404 2014; González-Sampériz et al. 2017). 405 The atmospheric reorganization following the YD led to a rapid resumption of the Atlantic 406 Meridional Overturning Circulation (AMOC) and a northwards return of the polar front to 50º-407 60ºN (Lane et al., 2013). This shifted the trajectory of the westerlies north to the Iberian 408 Peninsula, and thus weakening wind intensity and increasing humidity in Southern Europe. This 409 relative increase in moisture with respect to the previous scenario, the YD, was also reflected by 410 a decrease in the salinity of the lake water (Fig. 6) in Estanya and by an expansion of Juniper sp. 411 population in the watershed (Vegas-Vilarrubia et al., 2013; González-Sampériz et al., 2017). 412 During the remainder of the Holocene (7.5 to 0.6 cal. kyrs BP) the isotopic values of the lake 413 water averaged around 4.3±0.7‰ for  18 O and -1.5±2.9‰ for D, and showed less variability 414 than the Late Glacial owing to higher water level. The Holocene paleo-lake water values 415 recorded by gypsum are in accordance with modern  18 O and D of the lake water, which 416 indicates that environmental conditions were similar to present. During the Mid- Late-Holocene 417 (7.5 cal. kyrs BP to 0.6 cal. kyrs BP), atmospheric RH stabilized around ~70%. This value is 418 similar to the modern RH measured in the Lake Estanya region (annual mean of ~70-75%, 419 Perez-Bielsa, 2013). These results agree with previous reconstructions based on sedimentology 420 and geochemistry, which also show rather stable conditions similar to the present with short-421 lived abrupt hydrological fluctuations and an aridification trend after 4.5 to 4 cal. kyrs BP 422 (Morellón et al., 2009b). 423 424 6. Conclusions 425 We propose a new proxy for quantitative estimates of paleo-humidity. Analysis of GHW permits 426 the actual isotopic composition of paleo-waters to be determined, with little to no effect of 427 temperature. We couple triple oxygen and hydrogen isotopes in hydration water of lacustrine 428 gypsum and an Isotope Mass Balance model to quantify changes in RH in the past. Using Monte 429 Carlo simulations, the RH uncertainties derived from the input parameters to our model are 430 estimated. This can be as low as 3% (1) when the model is forced to match both the 17 O-excess 431 and d-excess of the paleo-water measured in GHW. 432 We apply this method to reconstruct the isotopic composition of paleo-waters of Lake Estanya 433 (NE Spain) and changes in atmospheric RH over the Late Glacial and Holocene periods (from 434 ~15 to 0.6 cal. kyrs BP). Our results indicate RH of 40-45% during the YD and increasing to 70-435 75% during the Mid-Late Holocene. This suggests that the mean RH in this region during the 436 past 7.5 cal. kyr BP was similar to present (RH~75%); however, the YD was characterized by 437 much drier conditions, with atmospheric RH ~30% lower than today. The southwards shift of the 438 Polar Front to ca. 42ºN during the coldest phases of the YD increased wind intensity and was 439 responsible for the minimum RH during this period. 440 The consistency of the results obtained from Lake Estanya with other proxies analyzed in this 441 lake and other regional paleoclimate records, demonstrates the reliability of isotopes in gypsum 442 hydration water as a tool for quantitative paleohydrological reconstructions in lake sediments. 443 Improving the analytical precision of triple oxygen isotope measurements in waters and better 444 understanding of the various parameters included in the model will reduce the uncertainties in 445 estimated RH. 446 ACKNOWLEDGMENTS 447 This research was supported by the ERC WIHM Project (#339694) to DAH. 448 449 References 450 Barkan, E., Luz, B., 2005. High precision measurements of 17 O/ 16 O and 18 O/ 16 O ratios in H2O, 451 Rapid Commun. Mass Spectrom. 19, 3737–3742. 452 Barkan, E., Luz B., 2007. Diffusivity fractionations of H2 16 O/H2 17 O and H2 16 O/H2 18 O in air and 453 their implications for isotope hydrology. Rapid Commun. Mass Spectrom. 21, 2999-3005. 454 Bateman, M.D., Díez-Herrero, A., 2001. The timing and relation of aeolian sand deposition in 455 central Spain to the aeolian sand record of NW Europe. Quat. Sci. Rev. 20, 779-782. 456 Brauer, A., Haug, G.H., Dulski, P., Sigman, D.M., Negendank, J.F.W., 2008. An abrupt wind 457 shift in western Europe at the onset of the Younger Dryas cold period. Nature Geosci. 1, 520-458 523. 459 Broecker, W.S., Andree, M., Wolfli, W., Oeschger, H., Bonani, G., Kennett, J., Peteet, D., 1988. 460 The chronology of the last Deglaciation: Implications to the cause of the Younger Dryas 461 Event. Paleoceanography 3, 1-19. 462 Cacho, I., Grimalt, J.O., Pelejero, C., Canals, M., Sierro, F.J., Flores, J.A., Shackleton, N., 1999. 463 Dansgaard-Oeschger and Heinrich event imprints in Alboran Sea paleotemperatures. 464 Paleoceanography 14, 698-705. 465 Criss, R.E., 1999. Principles of Stable Isotope Distribution, Oxford Univ. Press, Oxford, U. K. 466 Dongmann, G., Nürnberg, H.W., Förstel, H., Wagener, K., 1974. On the Enrichment of H2 18 O in 467 the Leaves of Transpiring Plants. Rad. and Environm. Biophys. 11, 41–52. 468 Escobar, J., Hodell, D. A., Brenner, M., Curtis, J. H., Gilli, A., Mueller, A. D., Anselmetti, F. S., 469 Ariztegui, D., Grzesik, D.A., Pérez, L., Schwalb, A., Guilderson, T.P., 2012. A ∼43-ka record 470 of paleoenvironmental change in the Central American lowlands inferred from stable isotopes 471 of lacustrine ostracods. Quat. Sci. Rev. 37, 92–104, 472 Evans, N.P., Turchyn, A., Gázquez, F., Bontognali, T.R.R., Chapman, H., Hodell, D., 2015. 473 Coupled measurements of  18 O and D of hydration water and salinity of fluid inclusions in 474 gypsum from the Messinian Yesares member, Sorbas Basin (SE Spain). Earth Planet. Sci. 475 Lett. 430, 499-510. 476 Eynaud, F., de Abreu, L., Voelker, A., Schönfeld, J., Salgueiro, E., Turon, J.-L., Penaud, A., 477 Toucanne, S., Naughton, F., Sánchez-Goñi, M.F., Malaizé, B., Cacho, I., 2009. Position of the 478 Polar Front along the western Iberian margin during key cold episodes of the last 45 ka. 479 Geochem., Geophy., Geosys. 10, Q07U05. 480 Fontes, J.C., Gonfiantini, R., 1967. Fractionnement isotopique de l’hydrogène dans l’eau de 481 cristallistion du gypse. Cr. Acad. Sci. d. Nat. 265, 4-6. 482 Frigola, J., Moreno, A., Cacho, I., Canals, M., Sierro, F.J., Flores, J.A., Grimalt, J.O., 2008. 483 Evidence of abrupt changes in Western Mediterranean Deep Water circulation during the last 484 50 kyr: a high-resolution marine record from the Balearic Sea. Quat. Int. 181 (1), 88–104. 485 García-Ruiz, J.M., Palacios, D., González-Sampériz, P., de Andrés, N., Moreno, A., Valero-486 Garcés, B., Gómez-Villar, A., 2016. Mountain glacier evolution in the Iberian Peninsula 487 during the Younger Dryas. Quat. Sci. Rev. 138, 16–30. 488 Gázquez, F., Mather, I., Rolfe, J., Evans, N.P., Herwartz, D., Staubwasser, M. Hodell D.A., 489 2015. Simultaneous analysis of 17 O/ 16 O, 18 O/ 16 O and 2 H/ 1 H of gypsum hydration water by 490 cavity ringdown laser spectroscopy. Rapid Commun. Mass Spectrom. 21, 1997-2006. 491 Gázquez, F., Evans, N.P., David. A. Hodell 2017a. Precise and accurate isotope fractionation 492 factors ( 17 O,  18 O and D) for water and CaSO4·2H2O (gypsum). Geochim. Cosmochim. 493 Acta, 198, 259-270. 494 Gázquez, F., Calaforra, J.M., Evans, N.P., David. A. Hodell 2017b. Using stable isotopes (δ 17 O, 495 δ 18 O and δD) of gypsum hydration water to ascertain the role of water condensation in the 496 formation of subaerial gypsum speleothems. Chem. Geol. 452, 34-46. 497 Gibson, J.J., Birks, S.J., Yi, Y., 2016. Stable isotope mass balance of lakes: a contemporary 498 perspective. Quat. Sci. Rev. 131, 316-328. 499 Gonfiantini, R., Fontes, J.C., 1963. Oxygen isotopic fractionation in the water of crystallization 500 of gypsum. Nature 200, 644-646. 501 González-Sampériz, P., Valero-Garcés, B.L., Moreno, A., Jalut, G., García-Ruiz, J.M., Martí-502 Bono, C., Delgado-Huertas, A., Navas, A., Otto, T., Dedoubat, J.J., 2006. Climate variability 503 in the Spanish Pyrenees during the last 30,000 yr revealed by the El Portalet sequence. Quat. 504 Res. 66, 38-52. 505 González-Sampériz, P., Aranbarri, J., Pérez-Sanz, A., Gil-Romera, G., Moreno, A., Leunda, M., 506 Sevilla-Callejo, M., Corella, J.P., Morellón, M., Oliva, B., Valero-Garcés, B., 2017. 507 Environmental and climate change in the southern Central Pyrenees since the Last Glacial 508 Maximum: A view from the lake records. Catena 149(3), 668-688. 509 Grauel, A.L., Hodell, D., Bernasconi, S.F., 2016. Quantitative estimates of tropical temperature 510 change in lowland Central America during the last 42 ka. Earth Planet. Sci. Lett. 438, 37-46. 511 Jasechko, S., Lechler, A., Pausata, F.S.R., Fawcett, P. J., Gleeson, T., Cendón, D. I., Galewsky, 512 J., LeGrande, A. N., Risi, C., Sharp, Z. D., Welker, J. M., Werner, M. and Yoshimura. K., 513 2015. Late-glacial to late-Holocene shifts in global precipitation δ 18 O. Clim. Past 11, 1375-514 1393. 515 Haese, B., Werner, M., Lohmann, G., 2013. Stable water isotopes in the coupled atmosphere-516 land surface model ECHAM5-JSBACH. Geosci. Model Dev. 6, 1463–1480. 517 Herwartz, D., Surma, J., Voigt, C., Assonov, S., Staubwasser, M., 2017. Triple oxygen isotope 518 systematics of structurally bonded water in gypsum. Geochim. Cosmochim. Acta 209: 254-266. 519 Hodell, D.A, Curtis, J.H., Brenner, M., 1995. Possible role of climate in the collapse of Classic 520 Maya civilization. Nature 375, 391–394. 521 Hodell, D.A., Benner, M., Curtis, J.H., 2005. Terminal Classic drought in the northern Maya 522 lowlands inferred from multiple sediment cores in Lake Chichancanab (Mexico). Quat. Sci. 523 Rev. 24, 1413–1427. 524 Hodell, D., Turchyn, A.V., Wiseman, C.J., Escobar, J., Curtis, J.H., Brenner, M., Gilli, A. 525 Mueller, A.D., Anselmetti, F., Aritzegui, D., Brown, E., 2012. Late Glacial temperature and 526 precipitation changes in the lowland Neotropics by tandem measurement of  18 O in biogenic 527 carbonate and gypsum hydration water. Geochim. Cosmochim. Acta 77, 352-368. 528 Kathayat, G., Cheng, H., Sinha, A., Spötl, C., R. Edwards, L., Li, H.Z., Yi, X., L., Ning, Y., Cai, 529 Y., Weiguo, L.L, Breitenbach, S.F.M., 2016. Indian monsoon variability on millennial-orbital 530 timescales. Sci. Rep. 6, 24374. 531 Landais, A., Barkan, E., Yakir, D., Luz, B., 2006. The triple isotopic composition of oxygen in 532 leaf water. Geochim. Cosmochim. Acta 70, 4105–4115. 533 Lane, C.S., Brauer, A., Blockley, S.P.E., Dulski, P., 2013. Volcanic ash reveals time-534 transgressive abrupt climate change during the Younger Dryas. Geology 41, 1251-1254. 535 Li, J., Li, M., Fang, X., Zhang, G., Zhang, W., Liu, X., 2017. Isotopic composition of gypsum 536 hydration water in deep Core SG-1, western Qaidam basin (NE Tibetan Plateau), implications 537 for paleoclimatic evolution. Global Planet Change. 155, 70-77. 538 Luz, B., Barkan E., 2010. Variations of 17 O/ 16 O and 18 O/ 16 O in meteoric waters. Geochim. 539 Cosmochim. Acta 74, 6276–6286. 540 Morellón, M., Valero-Garcés, B., Anselmetti, F., Ariztegui, D., Schnellmann, M., Moreno, A., 541 Mata, P., Rico, M., Corella, J.P., 2009a. Late Quaternary deposition and facies model for 542 karstic Lake Estanya (North-eastern Spain). Sedimentology 56, 1505-1534. 543 Morellón, M., Valero-Garcés, B., Vegas-Vilarrúbia, T., González-Sampériz, P., Romero, Ó., 544 Delgado-Huertas, A., Mata, P., Moreno, A., Rico, M., Corella, J.P., 2009b. Lateglacial and 545 Holocene palaeohydrology in the western Mediterranean region: The Lake Estanya record 546 (NE Spain). Quat. Sci. Rev. 28, 2582-2599. 547 Morellón, M., Valero-Garcés, B., González-Sampériz, P., Vegas-Vilarrúbia, T., Rubio, E., 548 Rieradevall, M., Delgado-Huertas, A., Mata, P., Romero, Ó., Engstrom, D., López-Vicente, 549 M., Navas, A., Soto, J., 2011. Climate changes and human activities recorded in the sediments 550 of Lake Estanya (NE Spain) during the Medieval Warm Period and Little Ice Age. J. 551 Paleolimnol. 46, 423-452. 552 Morellón, M., Benito, G., Moreno, A., González-Sampériz, P., Sánchez-Moya, Y., Pérez-Sanz, 553 A., Mata, P., Aranbarri, J., Sopeña, A., Valero-Garcés, B., 2014. Hydrological response to the 554 GS-1/Holocene transition in the Iberian Peninsula: environmental leads and lags, COST 555 INTIMATE MC Final workshop, Zaragoza (Spain). 556 Moreno, A., Gonzalez-Samperiz, P., Morellón, M., Valero-Garces, B.L., Fletcher, W.J., 2012a. 557 Northern Iberian abrupt climate change dynamics during the last glacial cycle: A view from 558 lacustrine sediments. Quat. Sci. Rev. 36, 139-153. 559 Ortiz, J.E., Torres, T., Delgado, A., Reyes, E., Llamas, J.F., Soler, V., Raya, J., 2006. Pleistocene 560 paleoenvironmental evolution at continental middle latitude inferred from carbon and oxygen 561 stable isotope analysis of ostracodes from the Guadix-Baza Basin (Granada, SE Spain). 562 Palaeogeogr. Palaeoclimatol. Palaeoecol. 240, 536–561. 563 Passey, B.H., Hu, H, Ji, H., Montanari, S., Li, S., Henkes, G.A., Levin, N.E. 2014. Triple oxygen 564 isotopes in biogenic and sedimentary carbonates. Geochim. Cosmochim. Acta 141,1–25. 565 Perez-Bielsa, C. 2013. Funcionamiento hidrogeológico de un humedal hipogenioco de origen 566 kárstico en las sierras marginas pirenaicas: Las lagunas de Estaña (Huesca). PhD Thesis. 567 Madrid, 419 pp. 568 Rach, O., Kahmen, A., Brauer, A., Sachse, D. 2017. A dual-biomarker appoach for 569 quantification of changes in relative humidity from sedimentary lipid D/H ratios. Clim. Past 570 Discuss. doi:10.5194/cp-2017-7. 571 Rasmussen, S.O., Seierstad, I.K., Andersen, K.K., Bigler, M., Dahl-Jensen, D., Johnsen, S.J., 572 2008. Synchronization of the NGRIP, GRIP, and GISP2 ice cores across MIS 2 and 573 palaeoclimatic implications. Quat. Sci. Rev., 27, 18-28. 574 Schoenemann, S.W., Schauer, A. J., Steig E. J. 2013. Measurement of SLAP and GISP  17 O and 575 proposed VSMOW-SLAP normalization for 17 O-excess. Rapid Commun. Mass Spec. 27, 576 582–590. 577 Sofer Z.,Gat J.R., 1975. The lsotopic composition of evaporating brines: effect of the isotopic 578 activity ratio in saline solutions. Earth Planet. Sci. Lett. 26, 179-186. 579 Steig, E.J., Gkinis, V., Schauer, A.J., Schoenemann, S.W., Samek, K., Hoffnagle, J., Dennis, 580 K.J., Tan, S.M., 2014. Calibrated high-precision 17 O-excess measurements using laser-current 581 tuned cavity ring-down spectroscopy, Atmos. Meas. Tech. 7, 2421-2435. 582 Surma, J., Assonov, S., Bolourchi, M.J., Staubwasser, M., 2015. Triple oxygen isotope 583 signatures in evaporated water bodies from the Sistan Oasis, Iran. Geophys. Res. Lett., 42, 584 8456–8462. 585 Torfstein, A., Gavrieli, I., Katz, A., Kolodny, Y., Stein, M., 2008. Gypsum as a monitor of the paleo-586 limnological-hydrological conditions in Lake Lisan and the Dead Sea. Geochim. Cosmochim., 72, 587 2491-2509. 588 Uemera, R., Barkan, E., Abe, O., Luz, B., 2010. Triple isotope composition of oxygen in 589 atmospheric water vapor. Geophys. Res. Lett. 37, L04402, doi:10.1029/2009GL041960. 590 Valero-Garcés, B., Morellón, M., Moreno, A., Corella, J.P., Martín-Puertas, C., Barreiro, F., 591 Pérez, A., Giralt, S., Mata-Campo, M.P., 2014. Lacustrine carbonates of Iberian Karst Lakes: 592 Sources, processes and depositional environments. Sedim. Geol. 299, 1-29. 593 Vegas-Vilarrúbia, T., González-Sampériz, P., Morellón, M., Gil-Romera, G., Pérez-Sanz, A., 594 Valero-Garcés, B., 2013. Diatom and vegetation responses to Late Glacial and Early 595 Holocene climate changes at Lake Estanya (Southern Pyrenees, NE Spain). Palaeogeogr., 596 Palaeoclimatol, Palaeoecol. 392, 335-349. 597 598 FIGURE CAPTIONS 599 Figure 1. Sensitivity of  18 O- 17 O-excess and  18 O-d-excess to different environmental 600 parameters during evaporation of a water body ( 18 O=-8‰D=-54‰, 17 O-excess=30 per meg 601 and d-excess=10‰) in partial equilibrium with atmospheric vapor. The isotopic composition of 602 a terminal lake has been model under different conditions of relative humidity (A and F), 603 temperature (B and G), wind (C and H) and degree of equilibrium between the atmospheric 604 vapor and the freshwater member (D and I). The isotopic compositions of water pools with 605 different ratios of Evaporation/Inflow (E/I), keeping the rest of parameter constant, are also 606 represented (E and J). 607 608 Fig. 2. A. Location of Lake Estanya. The red dashed line indicates the limits of the surface 609 catchment. Coring sites in Estanque Pequeño de Abajo are labelled. 610 Fig. 3.  18 O and D of rain and spring waters (green triangles), modern water from Lake Estanya 611 (red diamonds), and gypsum hydration waters (unfilled blue circles) ranging in age from 14.7 to 612 0.6 cal. kyrs BP. The isotopic values of paleo-lake waters (filled blue circles) were inferred using 613 isotopic fractionation factors between gypsum hydration water and the free solution (Gázquez et 614 al., 2017a). 615 616 Fig. 4. Results of the IMB model experiments (A. δ 18 O vs. 17 O-excess; B. δ 18 O vs. d-excess). 617 The colored diamonds represent the isotopic composition of Lake Estanya during the Holocene 618 (yellow), Preboreal (green), Younger Dryas (blue) and the Bølling-Allerød (pink) periods. The 619 model (blue lines and black dots) is tuned to fit the gypsum mother water compositions. 620 Environmental conditions for the different periods were simulated using the input parameters in 621 Table 1. The grey ellipses represent the uncertainty in the model derived from the tolerance 622 given for each input parameter. 623 624 Fig 5. Sensitivity experiments of the derived relative humidity to other major variables: 625 Evaporation/Input and the isotopic composition of the input. In both experiments data points 626 representative of the late-Holocene (blue markers) and the Younger Dryas (red markers) are used 627 to illustrate the potential biases in the derived relative humidity estimates: A. The derived 628 relative humidity shows a strong positive correlation to the assumed E/I in the range of 0.4-0.7 629 and the effect is greater at lower relative humidities (e.g. the YD). The effect is much weaker at 630 higher E/I in closed lake basins, such as Lake Estanya (as well as most other systems in which 631 evaporite mineral precipitation occurs); B. The derived relatively humidity with respect to the 632 assumed isotopic composition of the freshwater input. Here, relatively humidity shows a small 633 positive relationship with the isotopic composition of the freshwater input, which translates into 634 an error of between 3-4.6% in relative humidity for every 1‰ change in the freshwater water 635 input. 636 637 Fig. 6 Isotopic composition of Lake Estanya water, reconstructed atmospheric RH during the 638 Late Glacial-Holocene transition and the Holocene and comparison with other global and 639 regional paleoclimatic archives, including, from top to bottom: 1) NGRIP  18 O (Rasmussen et 640 al., 2008), 2) winter (blue line) and summer (red line) insolation at 42ºN, 3) Si/Al record in 641 marine core MD99-2343, offshore Minorca (Frigola et al, 2008); 4) Alboran Sea core MD95-642 2043 Sea Surface Temperature (SST) (Cacho et al., 1999); the isotopic composition of the paleo-643 lake water reconstructed from gypsum hydration water (panels 5, 6 and 7); 8) atmospheric RH 644 obtained from our 17 O-excess/d-excess model. The mean RH obtained from the 17 O-excess 645 model are represented as diamonds (see Fig. 4A). The RH results of the Monte Carlo simulation 646 in two scenarios are represented by the color shading. Scenario 1 (blue banding) in which all the 647 model parameters are held constant for all time periods and scenario 2 (orange banding) in which 648 the model parameters are modified in the Younger Dryas as described in the main text and 649 shown in Table 1. Previous palaeoenvironmental reconstruction of Lake Estanya based on 650 sedimentological and geochemical proxies, including 9)  13 C in organic matter, 10) a paleo-651 salinity proxy obtained from XRF analyses of the sedimentary sequence and, 11) relative lake 652 level reconstruction (0-10 stages) based on sedimentary facies (Morellón et al., 2009b). 653 654 Figure 1 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927290&guid=ff445cf8-b410-486e-a21a-a4fea2867ac9&scheme=1 Figure 2 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927291&guid=dbf9bca7-d444-4ea4-b5cf-99e9cc1a5b62&scheme=1 Figure 3 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927292&guid=133ef126-9f74-41a1-b4ee-e9baa055f19c&scheme=1 Figure 4 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927293&guid=6328dc81-66d8-44f4-9615-85bb0bea5483&scheme=1 Figure 5 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927294&guid=d45ea4bc-b39b-430c-8af2-5098c0e5f29e&scheme=1 Figure 6 Click here to download high resolution image http://ees.elsevier.com/epsl/download.aspx?id=927295&guid=fbe01c48-5e12-4051-a8e8-1e055f38b002&scheme=1 Supplementary material for online publication only Click here to download Supplementary material for online publication only: Supplementary material v. final 14-07-17.docx http://ees.elsevier.com/epsl/download.aspx?id=927378&guid=9ea1faea-a57e-4104-8c4a-e2d63eb239ee&scheme=1