Studio delle potenzialità delle tecniche di machine learning per valutare l'impatto delle variazioni del clima sull'ambiente e la salute umana

The thesis delves into the integration of machine learning with hydrology and public health to assess climate impact. It is structured in two main parts, each addressing a distinct aspect of this intersection. The first part focuses on improving precipitation simulation through a combined approach of statistical downscaling and machine learning. It specifically targets the shortcomings in predicting intermediate seasons and extreme precipitation events. This is achieved by enhancing Non-homogeneous Hidden Markov Models (NHMMs) with machine learning techniques, demonstrated in the case study of the Agro-Pontino Plain. Here, the model incorporates pre-processed atmospheric predictors to address NHMMs' limitations in representing extreme precipitation and seasonal variations, thereby providing a more accurate and robust simulation of rainfall patterns. In the second part, the thesis shifts its focus to the health impacts of climate change, particularly on cardiovascular diseases (CVD). Utilizing data from the Giovanni XIII Polyclinic in Bari, Italy, and incorporating weather and air quality data, a Random Forest machine learning model was developed to simulate trends in hospital admissions for CVD. The model's performance was rigorously evaluated, and the SHapley Additive exPlanations (SHAP) method was applied to ascertain the importance of various features. The study found that atmospheric pressure, minimum temperature, and carbon monoxide levels are critical factors influencing CVD-related hospitalizations, with atmospheric pressure being the most significant contributor. This research underscores the significance of integrating machine learning into the study of climate change's impacts on both environmental and health aspects. It highlights the critical role of climate variables in public health and provides a comprehensive framework for policymakers and healthcare professionals to mitigate the adverse effects of climate change on cardiovascular health. The thesis thus offers a multidimensional perspective on the climate crisis, combining advanced machine learning techniques with practical applications in environmental and health policy.