LEVERAGING THE BAYESIAN MODEL’S POSTERIOR DISTRIBUTIONS TO ENHANCE THE ACCURACY OF RAINFALL PREDICTIONS ACROSS UNSAMPLED LOCATIONS

Early flood detection is critical for mitigating the risk of flooding, particularly in regions prone to extreme weather events. Traditional spatial rainfall models often lack the temporal component necessary for accurate predictions, limiting their effectiveness. This study addresses these challenges by developing a Bayesian spatio-temporal model tailored to rainfall data across three distinct study areas: East Java (Indonesia), New South Wales (Australia), and the Red River Delta (Vietnam). The model integrates spatial and temporal data, enhancing its predictive capability and enabling more accurate forecasts. Specifically, we applied this model to rainfall data from these three regions, with the Inverse Distance Weighting (IDW) method used for interpola-tion. Rigorous statistical analysis and validation confirmed the model’s reliability in capturing moderate rainfall patterns, demonstrating strong predictive performance across varying spatial and temporal contexts. However, the model’s ability to predict extreme rainfall events remains limited, suggesting the need for further refinement. Despite this, the results highlight the model’s potential for practical applications, including early flood warning systems, enhanced irrigation planning, and improved water resource management.