Journal of Basic and Applied Research International

Background: Natural disasters such as typhoons, earthquakes, and tsunamis—exacerbated by climate change—pose significant global challenges, particularly for small island communities like Saipan. Given their increasing unpredictability and severity, there is an urgent need for AI-driven prediction and response systems capable of delivering accurate early warnings. Additionally, investigating the potential relationship between phytoplankton concentration changes and the occurrence of extreme weather events may contribute to the development of more proactive disaster management strategies.

Methods: This study developed and applied AI models to predict natural disasters using ten years (2014–2024) of data, including weather patterns, phytoplankton concentrations, and recorded events of typhoons, earthquakes, and tsunamis. The prediction system was designed for integration into Saipan’s disaster preparedness and recovery framework. Preprocessing steps included handling missing values using the K-Nearest Neighbors (KNN) imputation model, addressing outliers through linear regression, and normalizing the dataset. A random forest algorithm was employed to perform the predictions. Visualizations were generated to examine the correlation between natural disaster occurrences and variations in phytoplankton concentrations.

Results: The AI model demonstrated an overall prediction accuracy of approximately 99%, with an accuracy of 84% specifically for disaster-related events. While phytoplankton (chlorophyll) concentration alone showed weak correlations with individual weather variables such as temperature, precipitation, and wind speed, its predictive value increased when combined with other environmental features. Feature importance analysis revealed that climate-related variables—particularly wave height—were the most influential in predicting disaster occurrences. These results suggest that phytoplankton concentration, while not a strong standalone indicator, plays a meaningful role within a multi-variable prediction framework. The study highlights the importance of expanding environmental data collection and implementing real-time monitoring systems to improve forecasting precision. Integrating AI-based disaster prediction models into Saipan’s disaster response infrastructure could significantly enhance early warning capabilities, reduce recovery costs, and strengthen community resilience in the face of climate-driven natural hazards.