Abstract
Two modifications of the procedure for applying statistical corrections to the prediction equations discussed in Faller and Schemm (1977) are considered. In the first, the constant terms which normally appear in the regression equations are suppressed, leading to improved extended range STAT predictions. In the second, a noncentered advection term which gave exceptionally high correlations with one time step prediction errors is tested; however, we have been unable to use this type of term successfully in extended predictions. These results indicate the need for careful selection and testing of parametric terms for the proposed statistical corrections.
