We provide a globally distributed compilation of published surface temperature proxies for eight Cenozoic time periods that cover the range of paleoclimate states. The proxies have both a marine and terrestrial provenance and are compared to the annual temperature of the same location today. This data is then used to quantify long-term temperature changes on zonal and global levels. When coupled with recent estimates of atmospheric CO2 concentration, temperature data constrains the sensitivity of Earth's climate system to perturbation of the radiative balance, with possible implications for the future response to anthropogenic forcing.
The dataset consists of an excel file with eight sheets for the eight selected timeslices, namely,
• mid-Pliocene (3,0 - 3,3 Ma)
• late Miocene (7,2 - 11,6 Ma)
• mid-Miocene (14,7 - 17,0 Ma)
• early Miocene (20,3 - 23,0 Ma)
• early Oligocene (27,8 - 33,9 Ma)
• late Eocene (33,9 - 37,8 Ma)
• middle Eocene (42 - 46 Ma)
• early Eocene (48 - 55 Ma)

Most data were directly adopted from the primary sources. However, a number of published datasets were recalculated from the published geochemical proxy indices using the recent Bayesian core-top calibrations BAYSPAR (Tierney & Tingley, 2015), BAYSPLINE (Tierney & Tingley, 2018), BAYFOX (Malevich et al., 2019), BAYMAG (Tierney et al., 2019) and BAYMBT (Crampton-Flood et al., 2020). These are highlighted in blue in the dataset.
When more than one datapoint of a proxy is available for a given location, for example as part of a sea-surface temperature timeseries, then all datapoints falling within the bounds of our study timeslices were considered and the median was reported. Locations that are separated by less than 0.1 degrees in latitude and longitude were counted as a single locality.