Estuarine sediment deposition during wetland restoration: A GIS and remote sensing modeling approach

Abstract

Restoration is currently underway in the industrial salt flats of San Francisco Bay, California. Remote sensing of suspended sediment concentration and other GIS predictor variables were used to model sediment deposition within recently restored ponds. Suspended sediment concentrations were calibrated to reflectance values from Landsat TM 5 and ASTER satellite image data using three statistical techniques—linear regression, multivariate regression, and Artificial Neural Network (ANN) regression. Multivariate and ANN regressions using ASTER proved to be the most accurate methods, yielding r2 values of 0.88 and 0.87, respectively. Predictor variables such as sediment grain size and tidal frequency were used in the Marsh Sedimentation (MARSED) model for predicting deposition rates. MARSED results show a root mean square deviation (RMSD) of 66.8 mm (<1σ) between modeled and field observations. This model was applied to a pond breached in November 2010 and indicated that the pond will reach sediment equilibrium levels after 60 months of tidal inundation.