The evolution of 17O-excess in surface water of the arid environment during recharge and evaporation

This study demonstrates the potential of triple O-isotopes to quantify evaporation with recharge on a salt lake from the Atacama Desert, Chile. An evaporative gradient was found in shallow ponds along a subsurface flow-path from a groundwater source. Total dissolved solids (TDS) increased by 177 g/l along with an increase in δ18O by 16.2‰ and in δD by 65‰. 17O-excess decreased by 79 per meg, d-excess by 55‰. Relative humidity (h), evaporation over inflow (E/I), the isotopic composition of vapor (*RV) and of inflowing water (*RWI) determine the isotope distribution in 17O-excess over δ18O along a well-defined evaporation curve as the classic Craig-Gordon model predicts. A complementary on-site simple (pan) evaporation experiment over a change in TDS, δ18O, and 17O-excess by 392 g/l, 25.0‰, and −130 per meg, respectively, was used to determine the effects of sluggish brine evaporation and of wind turbulence. These effects translate to uncertainty in E/I rather than h. The local composition of *RV relative to *RWI pre-determines the general ability to resolve changes in h. The triple O-isotope system is useful for quantitative hydrological balancing of lakes and for paleo-humidity reconstruction, particularly if complemented by D/H analysis.


S1 Study Area and Samples
The Salar de Llamara is a salt flat in the hyperarid Atacama Desert of Chile. At W 69°37', S 21°16' there is a salt lake recharged only by groundwater. Local rainfall is absent. The nearest weather station to the Salar de Llamaraat Pozo Almonte, 110 km north of the lake siteshows an average annual precipitation < 1 mm/yr. On average, easterly winds prevail with a speed of 5 -15 m/s, but there is a thermally induced reversing diurnal pattern. Winds are strongest in the afternoon and blow with enough force to induce waves on the ponds and thus some extent of vertical mixing.
Average temperature (T) is 18 to 25 °C from October to March and 14 to 19 °C from April to September. Relative humidity is low and may fall below 20 %. During the field campaign (March, 3 rd to 7 th ) T and h were monitored on-site and showed a large intraday variation from 35 °C and 20 % h at the peak of the day, to 15 °C and 80 % during the night.
The ponds in the Salar de Llamara are shallow and between 0.5 and 1 m deep with the exception of pond 11, a sinkhole of about 2 m depth. Environmental protection measures have been enacted to counter a long-term decline in water level of the salar's remnant lake. A pumping station with three injection pipes branching off from a single supply-pipeline was constructed in late 2013 to refill the aquifer and prevent the lake from drying up. However, pumping had not yet been operational at the time of the sampling campaign in March 2014. Instead, groundwater water of low salinity was flowing through one of the injector pipes, rising under pressure from the subsurface aquifer.

S1.1 Natural Water Samples
During the campaign, we sampled groundwater and 11 ponds in the remnant lake system (Fig. S1).
Conductivity was measured on site, showing a range from 24.2 mS/cm to 174.1 mS/cm. Total dissolved solids (TDS, calculated from major element analysis) range from 16.4 g/l (pond 11) to 186 g/l (pond 1). Water from the aquifer was sampled from a pressure valve on the injector pipe with flowing water (sample 12, conductivity = 5.8 mS/cm and TDS = 4.2 g/l. In addition, we sampled a rare rainfall event near Antofagasta (260 km south of the lake) during the 2015 El Niño.

S1.2 Evaporation Experiment
A pan evaporation experiment was set up between the salar's ponds to investigate isotopic effects during evaporation without recharge for comparison. Three different waters -1.2 l of local tap water (Pica, 90 km northeast of the Salar de Llamara, TDS = 0.2 g/l), water from pond 8 (TDS = 22.5 g/l) and water from pond 1 (TDS = 186 g/l) -were filled into stainless steel pans (20 cm diameter). The experiment was conducted over three days and sampling was performed around 10:00 am and around 5:00 pm, starting in the evening of the first day after setting up the experiment in the morning. TDS of water from pond 8 increased to 59 g/l, that of pond 1 increased to 392 g/l over the course of the experiment. Evaporation of these saline waters was accompanied by precipitation of gypsum andin pond 1 waterof another soluble mineral, presumably mirabilite (Na2SO4 • 10H2O) as indicated by water chemistry and previously described 1 . The average fraction of water lost during the day was fd = 0.2 ± 0.05 and during the night was fn = 0.07 ± 0.02 (see Table   S5 for detailed information). Extrapolated evaporation rates to the year are 3,000 mm/yr (saline water) to 3,500 mm/yr (tap water). These are in good agreement with reported annual average rates of 3,500 to 4,000 mm/yr 2 . Total evaporation can be estimated using the open water surfaces (approx. 5,600 m 2 ) and the average annual evaporation (3,500 mm m/yr). This results in a total evaporation loss of approx. 18,000 m 3 /yr (50 m 3 /d). However, this estimate does not account for evaporation through the ground, which is clearly indicated by salt efflorescence between the ponds and the surroundings of the lake system.

S2 Parameterization
One indirectly estimated variable and two sensitive parameters affecting equations (5) -(7) are outlined in the following: These are the isotopic composition of atmospheric vapor ( * ), salinity effects on vapor pressure, i.e. the effective humidity (heff), and a wind turbulence correction (n).

S2.2 Salinity
Two aspects should be considered for calculating isotopic effects during evaporation of brines from equations (5) -(7): The 'classic' salt effect influences isotope activities 4-6 and the humidity effect describes the decrease of vapor pressure above the fluid with increasing salt content 7 .

S2.2.1 The "Classic" Salt Effect
The 'classic' salt effect (Γ) describes the fact that 18  have shown that NaCl has little effect on the isotopic activity of oxygen 11,12 , while others report a significant effect 13 . This chemical salt effect is small, however, for the single element system of triple oxygen isotopes. As such, we assume * = * , and neglect the 'classic' salt effect in our model calculations.

S2.2.2 Effective Humidity
In addition to a chemical salt effect, there is a physical effect. Dissolved salt raises water viscosity and lowers the vapor pressure. Slower evaporation at lower vapor pressure is parametrized by performing calculations with equations (2) and (5) -(7) with a higher humidity than actual, the effective relative humidity heff. There are two methods for calculation of heff 7,14 . In the first approach, ℎ = ℎ • 0 ⁄ where P0 is the vapor pressure of pure water and PS is the vapor pressure of the saline water 7 . The latter may be approximated from salinity with = 0 • (1 − 0.00407 − 0.000187 2 ) at 20 °C 14 . In a second approach heff is estimated using the measured density of the brines and Raoult's law 14 : Where ρ is the density of the brine and ρW is the density of pure water (taken as 0.9982 g/cm 3