Effects of Nonpoint Source Marsh Loading on Complex Estuaries

A foundational element in determining Total Maximum Daily Loads (TMDLs) is the determination of the relative impacts of point and nonpoint source impacts on the dissolved oxygen concentration (DO) of an impaired stream. It is not uncommon that the ultimate oxygen demand during a rain event can be greater than the effects of the fully permitted point source loading to the stream. Traditionally, the loading of oxygen consuming constituents is often estimated with a watershed model which is then coupled with a mechanistic model of the receiving stream. Calibrating coastal system applications of watershed models and mechanistic models to match the behavioral variability observed in actual field data is particularly difficult due to low watershed gradients, poorly defined drainage areas, chaotic forcing functions, and insufficient understanding of watershed and marsh process physics and chemistry. Data mining offers an alternative approach for analyzing and modeling tidal DO signals to quantify their responses to point and nonpoint source loadings. Data mining can be used to extract DO signal components whose forcing is by nonpoint loading caused by rainfall events and tidally entrained organics from marshes and mudflats. It has been successfully applied by the authors to South Carolina’s Cooper, Pee Dee, Savannah and Beaufort Rivers. At one gaging station on the Cooper River, rainfall was found to decrease DO concentrations at a rate of approximately 0.25 milligrams per liter (mg/L) per inch of rainfall. Similarly, it was found that specific conductance, water level, and tidal range, which indicate tidal forcing, modulate DO in the range of 3.1 mg/L. As confirmation of the process approach, rainfall impacts on the Beaufort River were found to be as high as 0.8 mg/L, whereas point source loads had an effect as high as 0.4 mg/L. This paper examines the approach of using data mining results to improve and expedite the calibration of mechanistic models for coastal applications and more accurately quantify the effects of nonpoint source load estimates.