Abstract
Since the 1970's, the frequency of vector-borne diseases such as Dengue, Chikungunya or Zika has significantly increased in the Pacific region. Understanding the factors and mechanisms underlying the spatio-temporal distribution of these diseases provides useful information regarding their control and prevention. In this thesis, we identified dengue spatio-temporal patterns and used modeling tools to identify the factors associated to an increased epidemiological risk at a regional scale (Pacific), a territorial scale (New-Caledonia), and a city scale (Noumea, the capital of New-Caledonia).Every five to seven years, dengue spreads over the entire Pacific as large epidemics caused by the introduction and regional diffusion of one of the four dengue virus serotypes. In New Caledonia, dengue has a seasonal epidemic pattern. The emergence of an epidemic requires specific climatic conditions. The identification of these conditions led to the implementation of an operational early warning system to predict dengue annual epidemic risk. Spatially, at the territorial scale, during epidemic years, high levels of viral circulation are found in areas with higher mean temperature and higher local population densities. Whether at the territorial scale or at the city scale, the spatial diffusion of the virus during epidemics caused by the re-emergence of the same serotype seems limited by the population immunity created by past epidemics. This thesis highlights the complexity and the multi-factorial aspect of vector-borne diseases, and discusses the usefulness of a multi-scale approach in modelling their epidemiology. Besides enhancing our understanding of dengue epidemiology over the Pacific area, we also developed a methodological framework that can be used in other geographical or epidemiological settings for the spatio-temporal analysis and modeling of epidemiological surveillance data.