A method of evaluating the transferability of streamflow information by regional regression analysis was applied in Washington to several streamflow variables. The annual mean and annual standard deviation were chosen to represent the development potential of the water resource, while the mean, standard deviation, and the 50-year recurrence interval of the annual flood series were chosen to represent flood potential. Low-flow characteristics were not used because of the inability to model them by regression analysis. Ephemeral streams in the vicinity of the Columbia Plateau were ignored for the same reason.
The results of the study indicate that the standard errors of estimate of the regression relations were good approximations of the medians of the Bayesian distributions of estimates of inaccuracy and that for the streamflow variables used in the study, little if any improvement can be expected in the regression relations by the collection of additional streamflow data. Improved transferability of streamflow information for these variables has a prerequisite of more accurate information-transfer models.
It is therefore recommended that the future streamflow-data network contain only those streamflow stations needed to provide data for (1) the design or operation of water-resources projects, (2) the investigation and study of water resources, and (3) monitoring long-term trends in streamflow.
