Development of regression equations to revise estimates of historical streamflows for the St. Croix River at Stillwater, Minnesota (water years 1910-2011), and Prescott, Wisconsin (water years 1910-2007)

A natural dam of glacial-era sediments at the confluence of the St. Croix and Mississippi Rivers forms Lake St. Croix, a riverine lake that comprises the lowest 25 miles of the St. Croix River. Historically, backwater effects from the Mississippi River prevented the use of traditional streamgages for collecting continuous streamflow data needed to calculate nutrient loads at the inlet to and outlet from Lake St. Croix at Stillwater, Minnesota and Prescott, Wisconsin, respectively. The development of index-velocity streamgages has enabled the measurement of continuous streamflow in backwater conditions using continuously measured velocities at the streamgage. Index-velocity streamgages were installed at Prescott, Wisconsin, and Stillwater, Minnesota, in 2007 and 2011, respectively.

Continuous daily mean streamflow data from the new index-velocity streamgages, long-term upstream streamgages, and tributary streamgages were used to (1) develop regression equations that improve estimates of historical streamflow at Stillwater and Prescott, (2) evaluate the accuracies of new and previous equations used to estimate historical streamflows, and (3) compute and evaluate revised estimates of historical streamflows for Stillwater for water years 1910–2011 and for Prescott for water years 1910–2007. The abilities of previous and newly developed regression equations to accurately estimate streamflows were evaluated using Nash-Sutcliffe Efficiency (NSE) values. The NSE values at Stillwater improved from 0.90 to 0.98, and the NSE values at Prescott improved from 0.77 to 0.94.

The new regression equations were used to calculate revised estimates of historical streamflows for Stillwater and Prescott starting in 1910 and ending when index-velocity streamgages were installed. Monthly, annual, 30-year, and period of record statistics were examined between previous and revised estimates of historical streamflows. The abilities of the new regression equations to estimate historical streamflows were evaluated by using percent differences to compare new estimates of historical daily streamflows to discrete streamflow measurements made at Stillwater and Prescott before the installation of index-velocity streamgages. Although less variability was observed between estimated and measured streamflows at Stillwater compared to Prescott, the percent difference data indicated that the new estimates closely approximated measured streamflows at both locations.