In the publication Mejía et al. (2017), for the calculation of diatom epsilon-p and pCO2 at 0.82 and 3.99 Ma we used the Uk'37 SST from the study of Lawrence et al. (2006), which used the Prahl et al. (1988) calibration. For the rest of the record, epsilon p and pCO2 data were calculated using the Uk'37 SST record calibrated using Müller et al. (1998) from the study of Herbert et al. (2016) and adapted to our age model (Shackleton et al., 1995).For all samples, including those at 0.82 and 3.99 Ma, the epsilon p and pCO2 data shown here are calculated using the Uk'37 SST calibration of Müller et al. (1998) from the study of Herbert et al. (2016), adapted to our age model (Shackleton et al., 1995). Changing the calibration used for the first two samples from Prahl et al. (1988) to Müller et al. (1998) (SST values taken from Herbert et al., (2016)), results in pCO2 estimates of up to 5.1 ppm higher than that reported in Mejía et al. (2017). The use of the Müller et al. (1998) calibration for the Uk'37 SST of the youngest sample would entail insignificant changes in the b* required to obtain a pCO2 of 229 ppmv (from 97, pCO2 of 233 ppmv to 95.5, pCO2 of 229 ppmv). This is not propagated through the remaining calculations because the reproducibility of the ice core pCO2 records is within this range of pCO2 variation (±3 ppmv). Data of the post 5 Ma alkenone concentrations were taken from the study of Lawrence et al. (2006). Data of Pre 5 Ma alkenone concentrations shown here are adapted to the age model of our diatom samples (Shackleton et al., 1995). The full record can be found in spreadsheet "Pre 5Ma alkenone and SST record".