A technique for determining relative ages of regions of the lunar surface using orbital photography was developed from a model of small-impact erosion. The erosion model relates the shape of a crater to the integrated flux of debris that has impacted the surface since that crater was fresh. The shape of the most modified crater of a particular diameter is thereby related to the relative age of the surface. Application of this analysis to orbital photography reveals that the major mare units vary in the accumulation of impacts by more than a factor of 3. Comparison of these data with crystallization ages determined from samples collected during the Apollo 11 and 12 missions indicates that the impact fluxes were decreasing during the stages of mare formation. An exponentially decaying flux for the last 3.5 b.y. with a half-life of 0.6 to 1.4 b.y. is compatible with the data. The data indicate (1) that the mare units sampled by the Apollo 11 and 12 missions represent major mare units of old and moderate age, respectively; (2) such major mare volcanism lasted on the order of 0.5 b.y. after the Apollo 12 basalts crystallized 3.3 b.y. ago; and (3) certain small localized volcanic features may be considerably younger (≳2 b.y.). The major mare unit at the Apollo 15 landing site (Hadley-Apennine) is expected to have a crystallization age of 3.5 ±0.1 b.y., whereas the crystallization age of the youngest unit in the Marius Hills region (a small mare ridge) is expected to have an age of 1 to 2 b.y., depending on the details of the impact flux history.