Assessment of temporal changes in malaria hazard modelling using surface water predictors

Abstract

The spread of malaria, worlds most deadly mosquito-borne disease is known to be directly linked to surface water through vector breeding. Despite this, many malaria prediction models do not use surface water as a predictor for malaria prevalence. Through improved resolution and easier remote sensing data availability, surface water-based predictors can prove to be an excellent way to create accurate malaria prevalence models. This study assesses the impact of yearly hydrological differences on the accuracy of a surface water-based malaria prevalence model and compares it to a precipitation-based malaria prevalence model. Using high-resolution remote sensing data to create yearly surface water maps, we compare a surface water-based malaria prevalence Boosted Regression Tree model with a precipitation-based malaria prevalence Boosted Regression Tree model for the period 2021-2023. We compared predictive performance and relative contribution of both surface water-based and precipitation-based predictors for both models. The predictive performance of the surface water model surpassed the precipitation model for all three years. High resolution surface water data proved to be a strong predictor of malaria. Additionally, our findings suggest that surface water predictors remain robust across varying hydrological conditions, highlighting their potential for determining malaria prevalence. As access to high-resolution satellite data increases, surface water-based models could provide more accurate, spatially refined malaria prevalence assessments, ultimately aiding public health interventions and vector control strategies.