Figure 1

Malaria risk maps developed in A) 1960s, B) 1990s and C) 2013. A) Malaria endemicity regions obtained from surveys between 1965–1967 [24]. B) Malaria risk map used in national malaria strategies and other MoH documents from 2005. Redrawn in colour from data from1967 [24], complimented by data collected in the early 2000 [44,45]. C) Population adjusted mean PfPR_2–10 (PAPfPR_2–10) in 2010 [77]. To compute PAPfPR_2–10 a Bayesian hierarchical space-time model was implemented through Stochastic Partial Differential Equations using Integrated Nested Laplace Approximations for inference [13]. The covariates were used together with the parasite rate data to predict malaria risk were temperature suitability index, precipitation, enhanced vegetation index and urbanization. Continuous maps of PfPR_2–10 maps were then predicted at 1 × 1 km. The continuous maps of PfPR_2–10 and matching population density grids projected for 2010 were then used to compute the number persons likely to be positive and the populations at risk by district at each 1 × 1 km PfPR_2–10 grid location. These were then used to compute PAPfPR_2–10.