Our understanding of past
climate
conditions largely
comes from paleo
climate proxies, su
ch as oxygen isotope ratios (δ18O
c) in marine
fossils. The marine δ18O
c signal primarily reflfle
cts a mixture of seawater temperature and oxygen isotopi
c composition of seawater (δ18Ow) at the time of
cal
cififi
cation. Knowledge of δ18Ow is
criti
cal for the interpretation of marine δ18O
c re
cords but remains poor for past hothouse
climates. Here, we
condu
ct water isotope-enabled simulations of the early Eo
cene using CO2 levels of 1×, 3×, 6×, and 9× the preindustrial value. We
cal
culate model δ18O
c using simulated δ18Ow and o
cean temperature, and make dire
ct
comparison with proxy re
cords. Model δ18O
c mat
ches the proxy values well for the early Eo
cene and Paleo
cene–Eo
cene Thermal Maximum with root-mean-squared errors approa
ching the standard error in individual re
cords. Eo
cene δ18Ow in the model exhibits strong variation depending on states of the hydrologi
cal
cy
cle and o
cean
cir
culation. Differen
ces in the mean δ18Ow between regions of net evaporation and pre
cipitation in
crease monotoni
cally with the magnitude of the net atmospheri
c moisture transport that
conne
cts them; however, this relationship breaks at the regional s
cale due to o
cean
cir
culation
changes. In parti
cular, an in
crease in o
cean ventilation brings more 18O-enri
ched deep water into the mixed layer, in
creasing sea-surfa
ce δ18Ow near the ventilation site and in
certain remote regions through fast upper o
cean
currents. δ18Ow variations and the linkage to both hydrologi
cal
cy
cle and o
cean
cir
culation bring
challenges for an a
ccurate interpretation of marine δ18O
c re
cords. Our study illustrates the value of using water isotope-enabled simulations and model-data
comparison for learning past
climate
changes. © 2020 Elsevier B.V. All rights reserved.