Log-Normal Streamflow Generation Models Reexamined

Data from several Australian rivers are used to examine three monthly log-normal streamflow generation models. The first model is based on Fiering’s skewed transformation equations which are applied to logs of monthly flows. The other two models are based on two- and three-parameter log normal distributions using moment transformation equations. Some deficiencies are noted in three procedures for computing storage size—the behavior approach, the sequent peak method, and Gould’s storage analysis. Length and number of traces are also studied. Large implied skew in the two-parameter model results in lower storage estimates than those found from the three-parameter model. Errors in the historical mean are more significant in relation to storage estimates than errors in serial correlation. The number of traces generated need only be sufficient to define storage mean and to indicate variability.