Triple oxygen isotopes in Atacama Desert waters since the late Miocene

Utilizing H and triple O isotopes, the main hydrologic variables from the Craig-Gordon model (CGM) of isotope evaporation from a lake, i.e. inflow composition (R_i), the atmosphere’s vapor composition (R_v), and relative humidity (h_r) can now be accurately constrained by measurements from differently evaporated subset lakes within the basin, if they fall on a single isotope evaporation trajectory in a diagram of ¹⁷O-excess or d-excess over δ¹⁸O. We demonstrate here, that this approach can be applied also to paleo-lakes by sampling subsets of lacustrine hydrous mineral deposits - e.g. gypsum (CaSO₄ • 2H₂O) – from the same geologic unit representing a narrowly constrained interval of time. This allows for the reconstruction of the above variables for the past. We conducted a proof-of-concept study in the Atacama Desert on modern and U-Pb dated paleo-gypsum lacustrine deposits. We tested the principles of the above approach on gypsum and lake water from the present-day Salar de Llamara. We verified signal preservation in a 1.8 Ma old – R_i constrained – marine lagoon gypsum outcrop situated on the tectonically uplifted Mejillones Peninsula that has been exposed to meteoric water for the last ~1 Ma. Finally, we applied the method to a 9 Ma old gypsum outcrop from the paleo-lake system of Tilliviche, which existed during the late Miocene / early Pliocene between ~ 11 Ma and 5 Ma. The CGM is applied to nine sub-samples with a ~ +13 to -10 per meg range in ¹⁷O-excess (~ +15 to -16 ‰ in d-excess). The model yields a paleo-R_i equal within model uncertainty to the present-day water flowing down Tilliviche ravine with a δ¹⁸O ~ -9 ‰ reflecting its high altitude source in the Andes. Paleo-R_v has a δ¹⁸O ~ -20 ‰, which is ~ 5 ‰ lower than the present-day atmosphere. Average modelled annual paleo-h_r is 66 % (44 to 84 % range), which is considerably more humid than at present (30 to 40 %). The data suggests that rainfall in the late Miocene Atacama Desert had an annual distribution equal to the present time and was likely as scarce. The more depleted past vapor composition can plausibly be explained only by higher subtropical rain-out prior to moisture advection into the desert. The higher paleo-h_r implies a much lower late Miocene evaporation rate, and must have been the prime cause of the lake’s existence. Thus, the late Miocene Atacama was likely already hyper-arid in terms of rainfall, but its atmosphere was more humid and less evaporative. The desert’s hyper aridity trend since the late Miocene fits with the global subtropical aridity trend beginning ~ 8 Ma ago and widely recorded by the expansion of drought-resistant C4 plants.