Coastal Water Quality Retrieval Based on Random Forest Coupled with Whale Optimization Algorithm using Sentinel-2 Data from Google Earth Engine

Monitoring water turbidity, as a vital indicator of water quality status, is crucial for the survival of aquatic ecosystems. Remote sensing data from the Sentinel-2 mission have been successfully used to retrieve turbidity information. To do so, machine learning models, such as Random Forest, have been employed. However, Random Forest works by simply averaging the predictions of each base model, which may not fully account for the performance of each model. In order to address this issue, this study proposed a Whale Optimization algorithm that combines the predictions of each base model in Random Forest to estimate water turbidity in Louisiana coastal waters. The proposed method was able to improve the performance of Random Forest by 0.08 in $R^{2}$ and more than 10% in Mean Absolute Percentage Error (MAPE).