Laserfiche WebLink
Q = a(Stage)b <br />(1) <br />where a and b were empirically derived. By taking the Loglo of both sides of the equation, the <br />equation can be put into a form where a and b can be estimated by linear regression: <br />Log10Q = a + bLoglo(Stage) <br />(2) <br />Equation 2 was modified slightly by subtracting a constant from the stage to reduce the range <br />of stage values compared to the mean stage (e.g., instead of using the elevation above sea level <br />for the stage, the stage was adjusted by subtracting a constant such as 1,899 ft). Mean daily <br />river stage values for each site were then regressed with the mean daily discharge (flow) for <br />the USGS river gaging station near each site. The USGS gage near Overton, NE (06768000) <br />was 3.6 miles upstream from the Elm Creek gage; the USGS gage near Kearney, NE <br />(06770200) was 9 miles upstream from the Rowe Sanctuary gage; and the USGS gage near <br />Grand Island, NE (06770500) was 11 miles down stream from the Crane Meadows gage. All <br />data for the period -of- record at each site were used, except the estimated flows for the USGS <br />gage and a few estimated river stages at each site. Data estimated by correlation with other <br />gages were routinely used to provide a continuous record when the data were missing due to <br />equipment malfunction or for other reasons (e.g., frozen river). The accuracy of estimated <br />data is seldom as good as actual data, however, and the USGS considers their missing data <br />poor. The time lao hetween the site and USGS gage was minimized by using the mean daily <br />float and stave. When the flow changed rapidly near the end of the day, however, the time lag <br />caused the relationsh!2 between the mean daily flow and stage to deviate from normal. Days <br />with an obvious time lag- thaefore, were also excluded. Ninety -five percent confidence belts <br />for predicting Q from a given stage were calculated using equation 16.27 from Zarr (1974). <br />PROC REG (SAS Institute Inc. 1985b) was used to calculate the regression equation, <br />and statistics from PROC REG and PROC MEANS (SAS Institute Inc. 1985a) were used to <br />calculate the 95 % confidence belts. The data were plotted using PROC GPLOT (SAS Institute <br />Inc. 1990b). <br />IVI <br />