Application of machine learning algorithms and Sentinel-2 satellite for improved bathymetry retrieval in Lake Victoria, Tanzania

• DOI: 10.1016/j.ejrs.2023.07.003
• Publication Date: 2023-07-16T01:46:21Z
• Authors: Makemie J. Mabula; Danielson Kisanga; Siajali Pamba
• Keywords: Gradient boosting

Estimating bathymetric information is vital for aquaculture and navigation applications. Free, high-resolution satellite imagery provides a cost-effective solution routine measurements. We tested six algorithms to retrieve water depth in the Mwanza Gulf of Lake Victoria using Sentinel-2 imagery: conventional Stumpf method, Random Forest (RF), Gradient Boosting Machine (GBM), Extreme (XGB), Neural Network (NNET), Support Vector (SVM). In-situ points collected via echo sounders were used train validate algorithms. Performance evaluation metrics included coefficient determination (R2), mean absolute error (MAE), root-mean-square (RMSE), spatial autocorrelation residuals. Among tested, model exhibited moderate performance with an R2 0.441, higher MAE (2.078 m), RMSE (2.964 m) values. The RF algorithm improved 0.957, lower (0.476 (0.823 m). GBM XGB achieved values 0.960 0.956, respectively, low (0.484 m GBM, 0.482 XGB) (0.795 0.830 NNET outperformed models, obtaining 0.963, lowest (0.438 (0.761 SVM demonstrated best 0.965, (0.403 (0.745 implying highest accuracy estimation. also showed stable generalization across different locations insignificant Therefore, recommended repetitive bathymetry calculations.