APPLICATIONS OF OPTIMIZATION MODELING ON A REGIONAL WATER MANAGEMENT MODEL: ACHIEVING TARGET SALINITY IN FLORIDA BAY, FLORIDA

The use of numerical modeling in evaluating regional water management leads to the simulation of various water delivery scenarios. The goal of producing a water delivery scenario is to analyze the effect the water management has on important factors such as water levels, flows, and salinities. It is common for these scenarios to be developed intuitively rather than analytically. In lieu of using a “trial and error” method for satisfying performance measures, an optimization technique may be used to provide a more precise result.

The U.S. Geological Survey's numerical modeling development of the coastal regions of Florida Bay and Everglades National Park (ENP) has led to an accurate representation of the surface-water and ground-water hydrology. The Southern Inland and Coastal System (SICS) model represents this area with a two-dimensional hydrodynamic surface-water model and a three-dimensional ground-water model, linked to represent the interaction of the two systems with salinity transport.

The SICS model is run within a parameter estimation program called UCODE. In this application, UCODE adjusts the regulated inflows to the ENP while SICS is run iteratively. The solution created by UCODE determines the response of salinities at a target location to these inflows. UCODE creates parameters defining input for the SICS model based on SICS model output statistics, with the objective of producing target salinities that meet ecosystem restoration goals. In order to apply this water-delivery-optimization technique to the SICS model area, allowable ranges in structure inflows and the salinity target criteria must be defined.

Preliminary results using two different parameterization methods illustrate the ability of the model to effectively improve salinity values in the target areas. In the final 173 days of the simulation, the original salinity's variance of 1.18 ppt2 is reduced to 0.47 ppt2 and 0.55 ppt2 by the two methods. Future model runs will include additional target areas for performance measures and longer simulation periods.