Salinization Simulation for the '1200-1500-1700-foot' Sands of the Baton Rouge Area, Southeastern Louisiana

摘要

Groundwater resources are vital for Louisiana's economic and demographic developments. Many freshwater aquifers underneath Baton Rouge are being contaminated by saltwater intrusion due to excessive groundwater withdrawals (Lovelace, 2007; Tsai, 2010). This study developed a three-dimensional SEAWAT model (Langevin et al., 2003) to simulate salinization in the "1200-foot" sand, the "1500-foot" sand, and the "1700-foot" sand in the Baton Rouge area. The study area shown in Figure 1 is cut across by the west-east trending Baton Rouge fault and the Denham Springs-Scotlandville fault. The Baton Rouge fault acts as a low-permeability leaky barrier, which allows saltwater encroachment from the south side of the fault. The major production wells for public supply are located in the area between two faults. The average pumping rate at the Lula pump station in is 7.03 million gallons per day and at the Government Street pump station is 1.59 million gallons per day. To develop the three-dimensional SEAWAT model, we first reconstructed the geological architecture of the Baton Rouge aquifer system using 288 electrical well logs. Then, we converted the geological architecture into a MODFLOW (Harbaugh 2005) computational grid. The MODFLOW grid consists of 45 layers. The "1200 foot" sand is from layer 1 to layer 20, and the "1500-1700-foot" sands are from layer 21 to layer 45, The Lula wells are screened from layers 28 to 33 and the Government Street wells are screened from layers 33 to 37. We developed a parallel version of the Covariance Matrix Adaptation-Evolution Strategy (CMA-ES) (Hansen et al., 2003) for high performance computing (HPC) in Louisiana supercomputers to calibrate the flow and transport models and estimate model parameters. The flow model was calibrated using 2805 groundwater head data to estimate permeability of the faults and conductivity and specific storage of the aquifers. Then, the MT3DMS (Zheng and Wang 1999) and SEAWAT models were constructed based on the developed MODFLOW modeL We used chloride data from 1975 to 2010 of 21 U.S. Geological Survey observation wells to calibrate the transport model and estimate porosity and dispersivities. The initial chloride concentration distribution was derived from the U.S. Geological Survey water quality database and from estimated salinity using spontaneous potential (SP) and deep electrical resistivity curves from 42 electrical logs in the area.