Skip Links

USGS - science for a changing world

Scientific Investigations Report 2011–5074

Prepared in cooperation with the U.S. Army Corps of Engineers–Savannah District

Simulation of Specific Conductance and Chloride Concentration in Abercorn Creek, Georgia, 2000–2009

By Paul A. Conrads, Edwin A. Roehl, Jr., and Steven R. Davie

Thumbnail image of report cover


The City of Savannah operates an industrial and domestic water-supply intake on Abercorn Creek approximately 2 miles from the confluence with the Savannah River upstream from the Interstate 95 bridge. Chloride concentrations are a major concern for the city because industrial customers require water with low chloride concentrations, and elevated chloride concentrations require additional water treatment in order to meet those needs. The proposed deepening of Savannah Harbor could increase chloride concentrations (the major ion in seawater) in the upper reaches of the lower Savannah River estuary, including Abercorn Creek.

To address this concern, mechanistic and empirical modeling approaches were used to simulate chloride concentrations at the city's intake to evaluate potential effects from deepening the Savannah Harbor. The first approach modified the mechanistic Environmental Fluid Dynamics Code (EFDC) model developed by Tetra Tech and used for evaluating proposed harbor deepening effects for the Environmental Impact Statement. Chloride concentrations were modeled directly with the EFDC model as a conservative tracer. This effort was done by Tetra Tech under a separate funding agreement with the U.S. Army Corps of Engineers and documented in a separate report. The second approach, described in this report, was to simulate chloride concentrations by developing empirical models from the available data using artificial neural network (ANN) and linear regression models. The empirical models used daily streamflow, specific conductance (field measurement for salinity), water temperature, and water color time series for inputs.

Because there are only a few data points that describe the relation between high specific conductance values at the Savannah River at Interstate 95 and the water plant intake, there was a concern that these few data points would determine the extrapolation of the empirical model and potentially underestimate the effect of deepening the harbor on chloride concentrations at the intake. To accommodate these concerns, two ANN chloride models were developed for the intake. The first model (ANN M1e) used all the data. The second model (ANN M2e) only used data when specific conductance at Interstate 95 was less than 175 microsiemens per centimeter at 25 degrees Celsius. Deleting the conductivity data greater than 175 microsiemens per centimeter removed the "plateau" effect observed in the data. The chloride simulations with the ANN M1 model have a low sensitivity to specific conductance (salinity) at Interstate 95, whereas the chloride simulations with the ANN M2 model have a high sensitivity to salinity at Interstate 95.

The two modeling approaches (Tetra Tech's EFDC model and the one described in this report) were integrated into a decision support system (DSS) that combines the historical database, output from EFDC, ANN models, ANN model simulation controls, streaming graphics, and model output. The DSS was developed as a Microsoft Excel™/Visual Basic for Applications program, which allowed the DSS to be prototyped, easily modified, and distributed in a familiar spreadsheet format. The EFDC and ANN models were used to simulate various harbor deepening scenarios. To accommodate the geometry changes in the harbor, the ANN models used the EFDC model-simulated salinity changes for a historical condition as input. The DSS uses a graphical user interface and allows the user to interrogate the ANN models and EFDC output.

Two scenarios were simulated using the Savannah Chloride Model DSS to demonstrate different input options. One scenario decreased winter streamflows to a constant streamflow for 45 days. Streamflows during the period January 1 to February 15 were set to a constant 3,600 cubic feet per second for the simulation period of October 1, 2006, to October 1, 2009. The decreased winter streamflow resulted in predictions of increased specific conductance by as much as 50 microsiemens per centimeter and chloride concentrations by as much as 4.8 milligrams per liter during the periods of decreased streamflows. The second scenario used EFDC output for a 4-foot deepening of the harbor and streamflow configurations to mitigate for salinity increases in the vicinity of an extensive freshwater tidal marsh. A 4-foot harbor deepening scenario was simulated for the 7-year period from January 2003 to October 2009. The ANN M2e model is more sensitive than the ANN M1e model to changes in specific conductance resulting from a 4-foot deepening and simulates chloride concentrations as high as 40 milligrams per liter. The ANN M1e model, which used all the data, simulated chloride concentration as high as 20.3 milligrams per liter.

First posted June 27, 2011

For additional information contact:
USGS South Carolina Water Science Center
Stephenson Center, Suite 129
720 Gracern Rd.
Columbia, SC  29210-7651
Phone: 803-750-6100

Part or all of this report is presented in Portable Document Format (PDF); the latest version of Adobe Reader or similar software is required to view it. Download the latest version of Adobe Reader, free of charge.

Suggested citation:

Conrads, P.A., Roehl, E.A., Jr., Davie, S.R., 2011, Simulation of specific conductance and chloride concentration in Abercorn Creek, Georgia, 2000–2009: U.S. Geological Survey Scientific Investigations Report 2011–5074, 46 p.





Purpose and Scope

Description of the Study Area

Previous Studies


Data-Collection Network

Limitation of the Datasets

Data Preparation

Characterization of Specific Conductance and Chloride Concentration

Simulation of Specific Conductance and Chloride Concentrations

Artificial Neural Network Models

Development of Specific Conductance and Chloride Models

Statistical Measures of Prediction Accuracy

Specific Conductance Models

Chloride Models

Development of a Decision Support System


Model Simulation Control and Streaming Graphics

Application of the Savannah Chloride Model Decision Support System

User-Defined Hydrograph

Inputs from Three-Dimensional Model Output

Summary and Conclusions

References Cited

Appendix 1. Savannah Chloride Model Decision Support System User's Manual

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo logo U.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: USGS Publications Team
Page Last Modified: Thursday, 10-Jan-2013 20:07:07 EST