A new technique is demonstrated for combining altimetric height (AH) and sea-surface temperature (SST) with in situ data to produce improved estimates of 0/800 m steric height (SH), heat content, and temperature variability. The technique uses a linear regression onto AH to construct an initial guess for the subsurface quantity. This guess is then corrected toward the in situ data creating an estimate with substantially less error than could be achieved using either data set alone. Inclusion of the SST data further improves the estimates and illustrates how the procedure can be generalized to allow inclusion of additional data sets. The technique is demonstrated over a region in the southwestern Pacific enclosing the Tasman Sea. Nine-year time series of heat storage and temperature variability, averaged over 4¿ latitude and longitude and 1 year in time, are calculated. The estimates have RMS errors of approximately 4.6 W/m2 in heat storage, 0.10¿C in subsurface temperature and 0.11¿C in surface temperature, and fractional errors of 20, 28, and 18%, respectively, relative to the total variance overall spatial and temporal scales considered. These represent significant improvements over previous estimates of these quantities. All the time series show strong interannual variability including the El Nino event of 1997. Application of these techniques on a global scale could provide new insight into the variability of the general circulation and heat budget of the upper ocean.