Machine learning-based prediction and forecasting of chlorophyll-a in the northern Indian Ocean using satellite data

Accurate prediction and forecasting of chlorophyll-a (Chl-a) concentrations are crucial for the sustainable management of marine ecosystems. While previous research has primarily focused on reconstructing and analyzing historical Chl-a data, limited efforts have been made to develop robust predictive models. This study addresses this gap by applying machine learning (ML) techniques to predict and forecast Chl-a concentrations in the northern Indian Ocean (NIO) using satellite-derived time series data from 1998 to 2022. Key environmental variables, including sea surface temperature (SST), sea surface height (SSH), and mixed layer depth (MLD), were integrated into the analysis due to their strong influence on primary productivity. Five ML models, XGBoost, long short-term memory (LSTM), Random Forest (RF), multi-layer perceptron (MLP), and support vector regression (SVR) were evaluated for their ability to handle complex temporal patterns and nonlinear interactions. Model performance was assessed using root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and the coefficient of determination (R 2 ). Among these models, XGBoost achieved the best performance, with an RMSE of 0.056, R 2 of 0.830, and MAPE of 0.078. LSTM, RF, and MLP also produced competitive results, whereas SVR was less effective due to its limitations in capturing data complexity. Forecasted Chl-a values revealed a declining trend in primary productivity, suggesting potential ecological shifts in the NIO. Overall, this study highlights the effectiveness of XGBoost for forecasting Chl-a using satellite data and demonstrate its potential as a scalable framework to inform climate change assessments, guide marine ecosystem monitoring, and support sustainable resource management.