The paper deals with the estimation of pedestrian risk exposure in urban areas. The approach proposed is based, on the one hand, on a spatial analysis technique called Space Syntax that characterizes the street integration - how streets are connected to each other - which gives rise to a natural movement of persons, and on the other hand, on land-use which acts as a multiplier or a divider of the original flows. Street integration is weighted by factors related to land use to better captures the heterogeneity of street-blocks. The method is applied to two radically different urban spaces: the northern periphery of the University of California at Berkeley (USA) and an area in downtown of Nantes (France). Five land-use factors are used to weight the integration: they express respectively the influence of residential areas, activities, public transportation, presence of sidewalks, and density of active frontage. Counting is performed at different locations and at different periods of the day (morning, lunch time, afternoon). Linear regression is performed between pedestrian volumes and the weighted and unweighted integration. Better correlation is obtained when using the weighted integration. No difference, in terms of correlation, is observed between the periods of the day. The study in Nantes shows that the street side has a significant effect on traffic volume. Discussions are made with respect to the differences between the cities of Berkeley and Nantes and their effect on traffic prediction
