A growing number of studies are concluding that the resilience of the Arctic sea ice cover in a warming climate is essentially controlled by its thickness. Satellite radar and laser altimeters have allowed us to routinely monitor sea ice thickness across most of the Arctic Ocean for several decades. However, a key uncertainty remaining in the sea ice thickness retrieval is the error on the sea surface height (SSH) which is conventionally interpolated at ice floes from a limited number of lead observations along the altimeter’s orbital track. Here, we use an objective mapping approach to determine sea surface height from all proximal lead samples located on the orbital track and from adjacent tracks within a neighborhood of 10s of kilometers. The patterns of the SSH signal’s zonal, meridional, and temporal decorrelation length scales are obtained by analyzing the covariance of historic CryoSat-2 Arctic lead observations, which match the scales obtained from an equivalent analysis of high-resolution sea ice-ocean model fields. We use these length scales to determine an optimal SSH and error estimate for each sea ice floe location. By exploiting leads from adjacent tracks, we can increase the SSH precision estimated at orbital crossovers by a factor of three. In regions of high SSH uncertainty, biases in CryoSat-2 sea ice freeboard can be reduced by 25% with respect to coincident airborne validation data. The new method is not restricted to a particular sensor or mode, so it can be generalized to all present and historic polar altimetry missions.