Standard regression models were used as a tool to evaluate the transferability of streamflow characteristics for the small-streams network in Missouri. Station records were divided into segments and tested for adequacy of record length and sample size for two physiographic regions. The standard error of estimate for each calibrated regression model, using increasing record lengths, was compared against the longest period of record in both tests. Results of the study indicate that an increase in the length of small-area streamflow records or additional sampling sites in the data network will not substantially improve the standard error of estimate of the regression model. Therefore, it is recommended that the current (1980) crest-stage network be reduced to a sample large enough to measure the long-term trend and provide information for future capabilities in the field of small-streams hydrology. It is also recommended that additional research be carried out to determine additional variables which could be used to predict floodflows and other types of regression models. (USGS)
