资源环境科技发展态势分析平台

Malaria is a climate sensitive disease that is causing rampant deaths in sub-Saharan Africa (SSA) and its impact is expected to worsen under climate change. Thus, pre-emptive policies for future malaria control require projections based on integrated models that can accommodate complex interactions of both climatic and non-climatic factors that define malaria landscape. In this paper, we combined Geographical Information System (GIS) and Bayesian belief networks (BBN) to generate GIS-BBN models that predicted malaria hotspots in 2030, 2050 and 2100 under representative concentration pathways (RCPs) 4.5 and 8.5. We used malaria data of children of SSA, gridded environmental and social-economic data together with projected climate data from the 21 Coupled Model Inter-comparison Project Phase 5 models to compile the GIS-BBN models. Our model on which projections were made has an accuracy of 80.65% to predict the high, medium, low and no malaria prevalence categories correctly. The non-spatial BBN model projection shows a moderate variation in malaria reduction for the high prevalence category among RCPs. Under the low prevalence category, an increase in malaria is seen but with little variation ranging between 4.6 and 5.6 percentage points. Spatially, under RCP 4.5, most parts of SSA will have medium malaria prevalence in 2030, while under RCP 8.5, most parts will have no malaria except in the highlands. Our BBN-GIS models show an overall shift of malaria hotspots from West Africa to the eastern and southern parts of Africa especially under RCP 8.5. RCP 8.5 will not expand the high and medium malaria prevalence categories in all the projection years. The generated probabilistic maps highlight future malaria hotspots under climate change on which pre-emptive policies can be based.