Spatial patterns and eco-epidemiological systems--part I: multi-scale spatial modelling of the occurrence of Chagas disease insect vectors.

Studies that explicitly and specifically take into account the spatial dimension within the study of eco-epidemiological systems remain rare. Our approach of modelling the spatial and/or temporal properties of the entomological and/or epidemiological data before their mapping with possible explanatory variables, objectively underline the significant patterns at different scales. The domiciliary and peri-domiciliary presence and abundance of juvenile and adult vectors of the Chagas disease (Triatoma sordida and Panstrongylus geniculatus) in Bahia state in northeast Brazil, has been modelled by automatically identifying significant multi-scale spatial patterns of the entomological data by the application and adoption of the spatial modelling methodology proposed by Dray et al. (2006) and based on principal coordinate analysis of neighbour matrices. We found that entomological data can be modelled by a set of eigenvectors that present a significant Moran's I index of spatial autocorrelation. The models for juvenile and adult vectors are defined by 28 and 32 eigenvectors that explain 82.3% and 79.9%, respectively, of the total data variances. The results support insect presence as the outcome both of a local scale "near-to-near" dispersal and an infestation from the wild, surrounding environment that produces a higher insect density at the village periphery.