The Clean Water Act requires the establishment of Total Maximum Daily Loads (TMDLs) for quantifying allowable pollutant loads for stream reaches in which the biological integrity of the stream is threatened. Sediment TMDLs in urban streams are particularly difficult to establish because they require (1) reliable measurement of sediment loads and (2) the ability to locate sediment sources. This research has attempted to address these challenges through a field study of North Peachtree Creek located in DeKalb County, Georgia, which has been sampled at the Century Boulevard crossing through automatic point sampling and depth-integrated sampling. Storm events from October 2003 through October 2004 provided a field record of sediment concentration and turbidity data over a wide range of storm events. Bed and bank sediment samples were collected for comparison with the point samples and depth-integrated samples. A methodology is presented whereby point sampling is used to calculate suspended sediment discharge and turbidity analysis is used to locate and characterize sediment sources. Point samples provide the boundary condition in the Rouse solution for the vertical distribution of suspended sediment to obtain suspended sediment discharge, which is then calibrated through comparison with depth-integrated sampling. The computer model HEC-RAS (U.S. Army Corps of Engineers, 1998) was applied to the stream reach to calculate the energy grade line slope throughout each storm event for input into the sediment discharge calculations. A favorable relationship between turbidity and suspended fine sediment was found at the sampling cross-section and, through comparison with bed and bank sediment samples, was used to identify the contribution of eroded bank sediment to the total sediment discharge.

Translational ecology and adaptive management strategies were incorporated into the Fordville Dam Nutrient Total Maximum Daily Load (TMDL) case study to determine if these two techniques were compatible to the North Dakota TMDL Program. A case study summary of the Fordville Dam Nutrient TMDL was discussed to provide contrast and comparison of the current TMDL program strategy and systematic improvements that could be made with the incorporation of translational ecology and adaptive management. Translational ecology is an effective way to bridge the information barrier through open communication between the stakeholders and scientists while creating a mutual learning experience. Adaptive management is beneficial to a TMDL implementation plan because it allows stakeholders and resource managers to become involved in management decisions and develop a better understanding of the ecosystem. Therefore, combining translational ecology and adaptive management would make the TMDL process more effective, through better communication and a flexible management plan.

Pursuant to the federal Clean Water Act (CWA), when a water body has been listed as impaired, Total Maximum Daily Loads (TMDLs) for the water quality constituents causing the impairment must be developed. A TMDL is the maximum daily mass flux of a pollutant that a waterbody can receive and still safely meet water quality standards. The development of a TMDL and demonstrating compliance with a TMDL requires pollutant load estimation. By definition, a pollutant load is the time integral product of flows and concentrations. Consequently, the accuracy of pollutant load estimation is highly dependent on the accuracy of runoff volume estimation. Runoff volume estimation requires the development of reasonable transfer functions to convert precipitation into runoff. In cold climates where a large proportion of precipitation falls as snow, the accumulation and ablation of snowpack must also be estimated. Sequential data assimilation techniques that stochastically combine field measurements and model results can significantly improve the prediction skill of snowmelt and runoff models while also providing estimates of prediction uncertainty. Using the National Weather Service's SNOW-17 and the Sacramento Soil Moisture Accounting (SAC-SMA) models, this study evaluates particle filter based data assimilation algorithms to predict seasonal snow water equivalent (SWE) and runoff within a small watershed in the Lake Tahoe Basin located in California. A non-linear regression model is then used that predicts suspended sediment concentrations (SSC) based on runoff rate and time of year. Runoff volumes and SSC are finally combined to provide an estimate of the average annual sediment load from the watershed with estimates of prediction uncertainty. For the period of simulation (10/1/1991 to 10/1/1996), the mean annual suspended sediment load is estimated to be 753 tonnes/yr with a 95% confidence interval about the mean of 626 to 956 tonnes/yr. The 95% prediction interval for any given year is estimated to range from approximately 86 to 2,940 tonnes/yr.