| Abstrak/Abstract |
The availability of shallow water bathymetry data is essential for support shipping safety, marine spatial planning, and conservation. Conventional bathymetric data acquisition generally requires complex instruments and is expensive. Remote sensing through multispectral image data allows the application of satellite-derived bathymetry (SDB) to obtain shallow water bathymetry data quickly and efficiently. Through this study, we tested the influence of seasonal variations across Indonesian waters: Northwest Monsoon (NWM), TS1 (Transition Season 1), Southeast Monsoon (SEM), and TS2 (Transition Season 2) on the accuracy produced by SDB through an empirical approach using Random Forest (RF) algorithm. The RF algorithm is applied to multispectral multiscenario images, which scale to surface reflectance (SR), deglint, and band ratio. The main variable is input depth data, which automatically divided into training and testing. The entire process is carried out on a cloud computing device called Google Earth Engine (GEE), thereby reducing processing residue and significantly saving processing time. We use two approaches to assessing accuracy: quantitatively using R2, MSE, and RMSE, and qualitatively descriptive using a 1:1 plot approach and underwater topographic profile via transect, which compares reference data with model depth. Overall, the resulting RMSE range is 0.86–1.7 m; for MSE, it is 0.84–2.9 m; and R2 is in the range of 0.63–0.86 m. This study found that seasonal variations have a systematic effect on accuracy, where NWM produces the lowest accuracy, which is thought to be due to atmospheric factors, and SEM TS2, respectively, produces the best accuracy. Through depth distribution, the SDB model is able to show maximum performance up to a total depth of 4 m. The underwater topographic profile shows that the overall scenario can replicate the depth well. This study provides comprehensive insight into the influence of seasonal variations on the accuracy of shallow-water bathymetric mapping |
