Observed Topex/Poseidon/Jason-1 (T/P/J) altimeter sea level (SL) data are used to improve an empirical parameterization of the temperature of subsurface water entrained into the mixed layer (Te). An inverse modeling method is first adopted to estimate Te from a SST anomaly (SSTA) equation using observed SST and simulated upper ocean currents from an intermediate ocean model (IOM). An empirical relationship between anomalies of Te (estimated) and SL (observed) from T/P/J altimeter is then constructed by utilizing a singular value decomposition (SVD) of their covariance. As compared with SSTA simulations using modeled SL data, the use of T/P/J SL data allows observed information about dynamic ocean adjustments to be built directly into the empirical Te model being used to parameterize the thermocline effect on SSTAs. The improved Te model leads to better SSTA simulations in the tropical Pacific. Cross validation is made to examine the sensitivity of the SSTA simulations to the period chosen for training the Te model. The proposed approach provides a new way to more effectively use T/P/J altimeter data in climate studies.