Changes in natural sand beaches induced by variations in incident waves were predicted by techniques of linear statistical estimation and empirical eigenfunction analysis. A 5-year set of measured beach profiles and wave statistics from southern California constituted the data base for this two-faceted statistical study. First, daily beach profile changes were predicted using four different spectral representations of the wave field. This profile changes were predictable using spectral representations of wave energy, radiation stress, energy flux, and wave steepness. Because of constraints on statistical reliability, a longer data set is required to select one of these as an optimal wave parameterization. Second, weekly beach profiles changes were predicted using weekly averaged wave characteristics. Weekly beach changes were predictable using weekly mean and maximum values of wave energy and wave height. The best predictor of those tested was the weekly mean wave energy. When combined with a longshore transport model, this onshore/offshore transport estimator should be applicable to other coastal regions with different beach and wave characteristics.