Simulation of Salinity in the Tidal Marshes in the Vicinity of the Savannah National Wildlife Refuge Using Artificial Neural Networks

Thumbnail Image
Conrads, Paul A.
Roehl, Edwin A.
Daamen, Ruby C.
Kitchens, Wiley M.
Associated Organization(s)
Supplementary to
The Savannah Harbor is one of the busiest ports on the East Coast of the United States and is located downstream from the Savannah National Wildlife Refuge, which is one of the Nation’s largest freshwater tidal marshes. The Georgia Ports Authority and the U.S. Army Corps of Engineers funded hydrodynamic and ecological studies to evaluate the potential effects of a proposed deepening of Savannah Harbor as part of the Environmental Impact Statement. These studies included a three-dimensional (3D) model of the Savannah River estuary system, which was developed to simulate changes in water levels and interstitial (or pore-water) salinity in the system in response to geometry changes as a result of the deepening of Savannah Harbor, and a marsh-succession model that predicts plant distribution in the tidal marshes in response to changes in the water-level and interstitial salinity conditions in the marsh. Beginning in May 2001, the U.S. Geological Survey entered into cooperative agreements with the Georgia Ports Authority to develop empirical models to simulate the water level and salinity of the rivers and tidal marshes in the vicinity of the Savannah National Wildlife Refuge and to link the 3D hydrodynamic river-estuary model and the marsh-succession model. Understanding freshwater inflows, tidal water levels, and specific conductance in the rivers and marshes is critical to enhancing the predictive capabilities of a successful marsh succession model. Data-mining techniques, including artificial neural network (ANN) models, were applied to address various needs of the ecology study and to integrate the riverine predictions from the 3D model to the marsh-succession model. ANN models were developed to simulate riverine water levels and specific conductance in the vicinity of the tidal marshes for the full range of historical conditions using data from the river gaging networks. ANN models also were developed to simulate the marsh water levels and interstitial salinities using data from the marsh gaging networks. Using the marsh ANN models, the continuous marsh network was indcasted to be concurrent with the long-term riverine network. The hindcasted data allow ecologists to compute hydrologic parameters—such as hydroperiods and exposure frequency—to help analyze historical vegetation data.
Sponsored and Organized by: U.S. Geological Survey, Georgia Department of Natural Resources, Natural Resources Conservation Service, The University of Georgia, Georgia State University, Georgia Institute of Technology
Date Issued
Resource Type
Resource Subtype
Rights Statement
Rights URI