Evidence for global mixing in real influenza epidemics

The spatiotemporal behaviour of the spread of influenza in France has been studied, and algebraic spatial correlations (with exponent α = −0.34) spanning the whole territory have been found to be present as soon as the number of reported cases begins to increase, about 15–25 weeks before the peak of the epidemic. This result is surprising, as one would expect long-range correlations, if any, only in the vicinity of the maximum incidence, whereas our observations suggest that there exists an underlying non-trivial spatial structure at the very beginning of the observed epidemic. The observed long-range correlations are in fact present in the spatial distribution of the population. Correlations in the number of cases normalized by local population density are characterized by α = 0. This suggests that the spread of the epidemic is statistically uniform in space over a complex substrate that already contains the observed long-range correlations