The electromagnetic environment in urban areas is growing increasingly complex. Sources of electromagnetic exposure like TV, FM, GSM, Wifi and others are spreading continuously and in the case of Wifi their geographical locations cannot be cataloged exhaustively anymore. Furthermore, the complexity of any highly urbanized environment and the lack of information about the dielectric properties of buildings lead to complex configuration so that a precise deterministic modeling of the electromagnetic exposure at any a given location of interest is probably out-of-reach.On the other hand there is a growing demand to assess the human exposure induced by these wireless communications. In a project between France Télécom R & D, Ecole des Mines and Supélec the application of geostatistical methods in this context is being explored.Geostatistics provides the right framework for setting up such exposure maps and its spatial statistical model yields an estimate of exposure as well as an associated error (De Doncker et al., 2006).The project consists of three phases: geostatistical evaluation of data generated by the numerical model EMF Visual (both in free space and with the addition of obstacles), statistical analysis of measurements performed in the area of the Quartier Latin in Paris and, finally, joint evaluation of an urban area both by statistical and deterministic numerical modeling.The paper reports about the third phase of this ongoing project, in which the spatial variation is modeled using the variogram, followed by a spatial regression known as kriging. The paper presents results about using a kriging algorithm that integrates numerical model output as an external drift.

No abstract