How does uncertainty influence spatial projections of Anopheles presence in Kenya?

Species distribution models (SDM) are becoming a widely used framework for studying distribution and risk of vector-borne diseases, particularly as a consequence of climate change (Gonzalez et al. 2010; Porretta et al. 2013; Rochlin et al. 2013). Malaria has been one of the most extensively studied vector-borne diseases (Minakawa et al. 2005; Ryan et al. 2006; Afrane et al. 2008; Mboera et al. 2010; Nath et al. 2012), and SDM output has been used by policy makers and various aid organizations to design and implement preventative malaria programs for areas that have been identified as current or future high risk (Gething et al. 2012; Hongoh et al. 2012; Cianci et al. 2015). However, these maps and models are often developed by epidemiologists or other medical researchers and therefore issues related to representing or exploring the uncertainty in the results have often been ignored (Lindsay et al. 1998; Levine et al. 2004). Many sources of uncertainty in model outputs have been identified in SDM research, ranging from data type or measurement level (e.g., presence-only vs. presence-absence, abundance), to statistical method, to subjective decisions related to mapping the results (e.g., threshold selected to discretize continuous output). This studies employs SDM to project the spatial distribution of four species of Anopheles (malaria-carrying mosquitoes) in Kenya, focusing on the representation of uncertainty and its propagation associated with aspects of the modeling methods and the data used.Geography and the Environmen