In a series of estimated flow statistics for an ungaged site, such as peak-flows at various recurrence intervals or flow-duration statistics, why do I sometimes get estimates for lower flows that are higher than estimates for higher flows?

The regression equations that are used to estimate flow statistics for ungaged sites were developed using computed streamflow statistics and measured basin characteristics for a selected group of streamgages. The regression equations have errors associated with them that can result in reversals of magnitude in a series of flow estimates. These errors occur because the stations that are used in the regression analysis are only a sample of the streams within the region, and the sample does not capture all of the variability in the streamflow statistics and basin characteristics that would be obtained if all streams within the region were gaged.

In addition, flow statistics and basin characteristics can’t be measured with 100 percent accuracy, and it is not possible to include in the regression equations all possible basin characteristics or other factors that cause flows to vary from site to site within the region. StreamStats provides one or more indicators of the accuracy of the estimates produced from regression equations.

Often, differences in the magnitudes of flow estimates within a series may be small (for example, the difference between the 85- and 90-percent flow), and the errors associated with the estimates may be relatively large. As a result, it is possible that an equation for the 90-percent flow will give an estimate that is larger than the estimate from an equation for the 85-percent flow, even though the 85-percent flow is a larger flow than the 90-percent flow. Usually in these cases, if prediction intervals are available for the estimates, the prediction intervals for the estimates will overlap each other.