WAVE HEIGHT FORECAST IN SANTOS BAY/SP: A HYBRID APPROACH OF GLOBAL NUMERICAL MODELS AND ARTIFICIAL INTELLIGENCE

Abstract

The city of Santos, located on the coast of São Paulo, Brazil, frequently experiences extreme metoceanographic events, such as storms and swells, exacerbated by climate change, which impact port and coastal safety. This study aims to enhance the forecasting of these events using machine learning techniques applied to global numerical models. Algorithms such as Random Forest Regressor, Extra Tree Regressor, and Gradient Boosting Regressor were employed, demonstrating high correlation (R > 0.95) and low errors (RMSE < 20 cm). However, the models struggled to predict more intense events, particularly waves exceeding 2 meters in height. Future work should focus on incorporating a longer historical period with more occurrences of intense events to improve model training and performance.

Keywords: Wave Height; Oceanographic Forecasting; Machine Learning