Online

Miake, Baxter G.

Civil and Environmental Engineering

2016

MS

Rivers deliver significant macronutrients and sediments to lakes that can vary substantially throughout the year. These nutrient and sediment loadings, exacerbated by winter and spring runoff, impact aquatic ecosystem productivity and drive the formation of harmful algae blooms. The source, extent and magnitude of nutrient and sediment loading can vary drastically due to extreme weather events and hydrologic processes, such as snowmelt or high flow storm events, that dominate during a particular time period, making the temporal component (i.e., time over which the loading is estimated) critical for accurate forecasts. In this work, we developed a data-driven framework that leverages the temporal variability embedded in these complex hydrologic regimes to improve loading estimates. Identifying the "correct" time scale is an important first step for providing accurate estimates of seasonal nutrient and sediment loadings. We use water quality concentration and associated 15-minute discharge data from nine watersheds in Vermont's Lake Champlain Basin to test our proposed framework. Optimal time periods were selected using a hierarchical cluster analysis that uses the slope and intercept coefficients from individual load-discharge regressions to derive improved linear models. These optimized linear models were used to improve estimates of annual and "spring" loadings for total phosphorus, dissolved phosphorus, total nitrogen, and total suspended loads for each of the nine study watersheds. The optimized annual regression model performed ~20% better on average than traditional annual regression models in terms of Nash-Sutcliffe efficiency, and resulted in ~50% higher cumulative load estimates with the largest difference occurring in the "spring". In addition, the largest nutrient and sediment loadings occurred during the "spring" unit of time and were typically more than 40% of the total annual estimated load in a given year. The framework developed here is robust and may be used to analyze other units of time associated with hydrologic regimes of interest provided adequate water quality data exist. This, in turn, may be used to create more targeted and cost-effective management strategies for improved aquatic health in rivers and lakes.