Our understanding of past c
limate conditions
large
ly comes from pa
leoc
limate proxies, such as oxygen isotope ratios (δ18Oc) in marine fossi
ls. The marine δ18Oc signa
l primari
ly ref
lf
lects a mixture of seawater temperature and oxygen isotopic composition of seawater (δ18Ow) at the time of ca
lcifification. Know
ledge of δ18Ow is critica
l for the interpretation of marine δ18Oc records but remains poor for past hothouse c
limates. Here, we conduct water isotope-enab
led simu
lations of the ear
ly Eocene using CO2
leve
ls of 1×, 3×, 6×, and 9× the preindustria
l va
lue. We ca
lcu
late mode
l δ18Oc using simu
lated δ18Ow and ocean temperature, and make direct comparison with proxy records. Mode
l δ18Oc matches the proxy va
lues we
ll for the ear
ly Eocene and Pa
leocene–Eocene Therma
l Maximum with root-mean-squared errors approaching the standard error in individua
l records. Eocene δ18Ow in the mode
l exhibits strong variation depending on states of the hydro
logica
l cyc
le and ocean circu
lation. Differences in the mean δ18Ow between regions of net evaporation and precipitation increase monotonica
lly with the magnitude of the net atmospheric moisture transport that connects them; however, this re
lationship breaks at the regiona
l sca
le due to ocean circu
lation
changes. In particu
lar, an increase in ocean venti
lation brings more 18O-enriched deep water into the mixed
layer, increasing sea-surface δ18Ow near the venti
lation site and in certain remote regions through fast upper ocean currents. δ18Ow variations and the
linkage to both hydro
logica
l cyc
le and ocean circu
lation bring cha
llenges for an accurate interpretation of marine δ18Oc records. Our study i
llustrates the va
lue of using water isotope-enab
led simu
lations and mode
l-data comparison for
learning past c
limate
changes. © 2020 E
lsevier B.V. A
ll rights reserved.