Determination of The Best Methodology for Bathymetry Mapping using Spot 6 Imagery: A Study of 12 Empirical Algorithms

For the past four decades, many researchers have published a novel empirical methodology for bathymetry extraction using remote sensing data. However, a comparative analysis of each method has not yet been done. Which is important to determine the best method that gives a good accuracy prediction. This study focuses on empirical bathymetry extraction methodology for multispectral data with three visible band, specifically SPOT 6 Image. Twelve algorithms have been chosen intentionally, namely, 1) Ratio transform (RT); 2) Multiple linear regression (MLR); 3) Multiple nonlinear regression (RF); 4) Second-order polynomial of ratio transform (SPR); 5) Principle component (PC); 6) Multiple linear regression using relaxing uniformity assumption on water and atmosphere (KNW); 7) Semiparametric regression using depth-independent variables (SMP); 8) Semiparametric regression using spatial coordinates (STR); 9) Semiparametric regression using depth-independent variables and spatial coordinates (TNP), 10) bagging fitting ensemble (BAG); 11) least squares boosting fitting ensemble (LSB); and 12) support vector regression (SVR). This study assesses the performance of 12 empirical models for bathymetry calculations in two different areas: Gili Mantra Islands, West Nusa Tenggara and Menjangan Island, Bali. The estimated depth from each method was compared with echosounder data; RF, STR, and TNP results demonstrate higher accuracy ranges from 0.02 to 0.63 m more than other nine methods. The TNP algorithm, producing the most accurate results (Gili Mantra Island RMSE = 1.01 m and R2=0.82, Menjangan Island RMSE = 1.09 m and R2=0.45), proved to be the preferred algorithm for bathymetry mapping