Monitoring and Prediction of Seawater pH Variations in Sinop Coastal Waters Using Machine Learning-Based Models

Acidification, a critical indicator of global climate change, has a profound and irreversible impact on biodiversity hotspots. These effects ultimately lead to ecosystem collapse and profound changes in known life forms. In the Black Sea, which is isolated from the open oceans, acidification may cause severe ecosystem damage. Recently, monitoring acidification in marine environments has gained significant attention. This study aims to monitor changes in the pH levels of seawater in the offshore region of Sinop (Black Sea, Türkiye) and to evaluate the predictive capability of the Gaussian Process Regression (GPR) model, a machine learning method, in forecasting these changes. For the study, daily seawater data for the Sinop region of the Black Sea, obtained via the Copernicus satellite, were used for the period between 1 June 2022 and 28 March 2025. The Copernicus Marine Environment Monitoring Service (CMEMS) publishes annual reports on the health of the oceans and regional seas of Europe, contributing to awareness of the state and transformations of marine environments. In the analysis section of this study, error metrics such as MAPE, RMSE, and R² were utilized to assess the accuracy and predictive performance of the models. During model development, variables actively influencing pH levels, such as sea surface temperature and partial pressure of carbon dioxide (spCO₂), were taken into account. Among the tested kernel functions, the Rational Quadratic kernel achieved the best performance with MAPE = 0.0177, RMSE = 0.0003, and R² = 0.9731, indicating a highly reliable predictive capability for seawater pH levels. The results obtained provide significant insights into the regional-scale monitoring of seawater chemistry and the early detection of potential environmental changes.