Integrating Climate and Environmental Data with Bayesian Models for Malaria Prediction

Abstract

Malaria remains a notable public health challenge in endemic regions, with an estimated 263 million cases and 579,000 malaria-related deaths globally in 2023. Climate and environmental factors, such as temperature, rainfall, and the Normalised Difference Vegetation Index (NDVI), play a crucial role in malaria transmission. While statistical models aid in malaria prediction, Bayesian methods remain underutilised despite their ability to incorporate prior knowledge into predictive models. The major contribution of this study is to develop a Bayesian malaria prediction model incorporating climate and environmental data. Both objective and subjective prior distributions are evaluated to determine their effectiveness in improving model performance. The results indicate that a subjective prior outperforms other priors. Additionally, Ehlanzeni (Mpumalanga), Vhembe and Mopani districts (Limpopo) are identified as high-risk malaria regions. The findings suggest that malaria transmission peaks in summer and autumn, particularly in areas where temperatures during the night range from 12°C-20°C, rainfall is moderate (100–200 mm), and NDVI exceeds 0.6. Malaria risk intensifies following months of accumulated rainfall, creating optimal mosquito breeding conditions. These insights may assist malaria control programmes in developing targeted interventions, such as early warning systems and vector management strategies. Future research will explore Bayesian machine learning for malaria prediction.