Shallow Water Depth Estimation with Geographically Weighted Regression Model Using Multispectral Satellite Imagery

Abstract

Shallow water depth estimation using passive remote-sensing method is an attractive alternative as it provides a time- and cost-effective solution to water depths estimation. Among all the models that have been presented in the literature for Satellite Derived Bathymetry (SDB), the algorithm proposed by Lyzenga et al. (2006) is still the most popular one, due to its physically intuitive nature. The common practice adopted in previous attempts on the Lyzenga et al. (2006) model has been to calibrate a single set of parameters using global regression model. But it's well known that the global inversion model's optical uniformity assumption is unrealistic and not suitable for coastal and inland water bodies where the bottom type and water quality vary spatially. To address this inadequacy, we use geographically adaptive approach of Lyzenga et al. (2006) model (local model) that takes into account local factors in determining model parameters in order to better estimate bottom depth. The accuracy assessment was based on the coefficient of determination R2 and the Root Mean Square Error (RMSE). Results demonstrate that the local inversion model performs well in estimating bathymetry of the shallow waters of the Bandama estuary, showing R2 of 0.993 and RMSE of 0.51 m. Thus, the results obtained indicate that the local inversion model may be able to provide an estimate of bathymetry for many coastal areas in Côte d’Ivoire.