Flow reversal events and statistical modeling of flow dynamics of hypersaline water across a constructed causeway, Great Salt Lake, Utah, USA

The hypersaline Great Salt Lake (GSL), located in the western United States, is divided into a north arm (Gunnison Bay) and a south arm (Gilbert Bay) by a 29-km earth-filled railroad causeway completed in 1959. Flow between the two bays is restricted to an 88-m wide breach at the western end of the causeway known as the GSL Breach. A gauging station at the GSL Breach monitors flow across the causeway. Flow across the GSL Breach is dynamic, ranging from -166 m3/s to 340 m3/s over the 5-year analysis period (2009-2013). The prominent flow direction is from south-to-north reported as positive discharge and flow from north-to-south is reported as negative discharge. Negative discharge events at the GSL Breach are referred to as flow reversal events (FRE). During the 5-year period of analysis, a total of 1510 FRE were recorded at the GSL Breach gauging station, with an average of 302 events occurring each year. Several FREs were short in duration, lasting 15-minutes or less. The longest event recorded at the GSL Breach was 1950-minutes (32.5 hours). FRE are most common when the difference between the water surface elevations of Gilbert and Gunnison Bay is at a minimum (September through December). Longer duration FRE are most common during the cool seasons (September to May) when large storm systems cross over the GSL Basin. To determine the variables influencing discharge and FREs at the GSL Breach, we applied several predictor variables to a generalized additive model (GAMs). GAMs were used to develop an effective predictive model for discharge at the GSL Breach using smoothing functions of predictor variables measured at various hydrologic and meteorological stations. We assessed the use of GAMs for accurate prediction of discharge by statistical analysis, residual plots, and time-series analysis using observed daily values at the GSL Breach gauging station. Models using wind gust, wind direction, lake elevation, and stream stage as predictor variables produce valid explanatory results for discharge at the GSL Breach. GAMs proved to be an effective method for capturing the rapid changes in discharge observed at the GSL Breach, suggesting the sensitivity of the GAM is of sufficient resolution for understanding the hydrodynamics of the GSL. To improve model accuracy, additional lake level monitoring equipment should be installed on the north and south side of the causeway and a weather station at the center of the GSL causeway is needed.