The channel geometry of 48 gaged streams in California where mean annual runoff is known was studied in 1967 and 1968. The analyses show that the mean annual runoff is related to selected dimensions of channel geometry. The width and the average depth of the cross section between bars or beams can be used to estimate annual runoff from ungaged streams. Separate relations are needed for perennial and ephemeral streams. The analyses also showed that it is better to measure several cross sections, compute the discharge for each cross section, and average these discharges to obtain the discharge for the site. A 10-year period, 1958-67, was analyzed to determine if the channel dimensions were affected by recent hydrologic or climatic events. It was determined that the computed runoff represented a long-term mean; that is, the standard error of estimate was less for the regression using the runoff for the period of record rather than for the 10year period.