Résumé
This paper deals with the characterization of the landscape composition and structure, by means of high resolution remote sensing, in order to explain the presence and the spatial distribution of Chagas disease vectors in a locality of the Iraquara municipality (Bahia, Brazil) where the disease is endemic. At a local scale, many other parameters can influence Chagas vector presence and dispersion dynamics, particularly dwelling, peri-domiciliary space and human behavior characteristics. In this study, these factors have been characterized by means of field works and inhabitant inquiries. Factorial analysis of mixed groups is then proposed to jointly analyze these numerous and heterogeneous data. Results shows that i) P. geniculatus species is associated with domiciliary units (DUs) situated far from the village center, with open peri-domiciliary spaces, particular landscape characteristics ; ii) presence of nymph(s) and adult vector(s) are clearly not associated to badly-kept houses and poor building materials; iii) one DU can be responsible for the dispersion of T. sordida vector in the adjacent DUs and iv) vector(s), especially adult(s), seems to be associated with the presence of hen, of henhouse and of lighting outside. This work shows the pertinence of the methodology and is a first step towards the definition of indicators of vector presence, density and spatial distribution that would permit to support the actions of the national Chagas disease control program and to set up alert systems. Keywords: remote sensing, landscape characterization, factorial analysis of mixed groups, chagas disease sensoriamento remoto, caracterização da paisagem, análise fatorial de grupos mistos, Doença de Chagas. 1. Introduction Environment plays a key role in the (re-)emergence and endemicity of many infectious vector-borne diseases. In fact, presence, density and spatial distribution of vectors and animal hosts depend on biotic and abiotic conditions. In this context, the importance of landscape composition (number and types of patches) and configuration (spatial relationships among patches) to disease dynamics is nowadays being explored [Ostfeld, Glass e Keesing 2005]. High resolution remote sensing appears to be an appropriate tool to deal with the relationships between environment and health by means of landscape epidemiology approach. In fact, remote sensing has been widely used for health issues monitoring, disease (re-)emergence explanation and prediction, risk map elaboration. However, effective contributions of remote sensing to health problematics are mitigated [Herbreteau et al. 2007] : high spatial resolution is rarely used, multispectral imagery is almost absent, NDVI is used in the majority of the studies without research of specific and potentially more pertinent indexes, and spatial scales at which studies are performed are not always appropriate to health issues. Another limit of the existing studies is the fact that they rarely jointly analyze the different factors that have a possible impact on the (re)-emergence and on the endemic character of the disease. Some studies try to overcome these limits by using high resolution imagery and developping specific indexes to identify and