A regional low-flow survey of small, perennial streams in western Washington was initiated by the Northwest Indian Fisheries Commission (NWIFC), NWIFC-member tribes, and Point-No-Point Treaty Council in cooperation with the U.S. Geological Survey in 2007 and repeated by the tribes during the low-flow seasons of 2008–09. Low-flow measurements at 63 partial-record and miscellaneous streamflow-measurement sites during surveys in 2007–09 are used with concurrent flows at continuous streamflow-gaging stations (index sites) within the U.S. Geological Survey network to estimate the low-flow metric Q7,10 at each measurement site (Q7,10 is defined as the lowest average streamflow for a consecutive 7-day period that recurs on average once every 10 years). Index-site correlation methods for estimating low-flow characteristics at partial-record sites are reviewed and an empirical Monte Carlo technique is used with the daily streamflow record at 43 index sites to determine the error and bias associated with estimating the Q7,10 at synthetic partial-record sites using three methods: Q-ratio, MOVE.1, and Base-Flow Correlation. The Q-ratio method generally has the lowest error and least amount of bias for 170 scenarios, with each scenario defined by the number of concurrent flow measurements between the partial-record and index sites (ranging from 4 to 20) and the combination of basin attributes used to select the index site. The root-mean square error for the Q-ratio method ranged from 70 to 118 percent, depending on the scenario. The scenario with the smallest root-mean square error used four concurrent flow measurements and the basin attributes: basin area, mean annual precipitation, and base-flow recession time constant, also referred to as tau (τ).
Regional low-flow regression models for estimating Q7,10 at ungaged stream sites are developed from the records of daily discharge at 65 continuous gaging stations (including 22 discontinued gaging stations) for the purpose of evaluating explanatory variables. By incorporating the base-flow recession time constant τ as an explanatory variable in the regression model, the root-mean square error for estimating Q7,10 at ungaged sites can be lowered to 72 percent (for known values of τ), which is 42 percent less than if only basin area and mean annual precipitation are used as explanatory variables. If partial-record sites are included in the regression data set, τ must be estimated from pairs of discharge measurements made during continuous periods of declining low flows. Eight measurement pairs are optimal for estimating τ at partial-record sites, and result in a lowering of the root-mean square error by 25 percent. A low-flow survey strategy that includes paired measurements at partial-record sites requires additional effort and planning beyond a standard strategy, but could be used to enhance regional estimates of τ and potentially reduce the error of regional regression models for estimating low-flow characteristics at ungaged sites.