We introduce in this paper a soundscape simulator called SimScene, designed to be used as an experimental tool to characterize the mental representation of sound environments. The soundscape simulator allows a subject to generate a full sonic environment by sequencing and mixing sound elements, and manipulating their sound level and time positioning. To make the simulation process effective, SimScene has not be designed to manipulate individual parameters of individ-ual sounds, but to specify high-level parameters for whole classes of sounds, organized into a hierarchical semantically structured dataset. To avoid any linguistic bias, a listening oriented interface allows subjects to explore the dataset with-out any text written help. The entire software is developed in Javascript using the standard Web Audio technology, and is thus fully supported by most modern web browsers. This fact should allow experimenters to adopt a crowdsourcing approach to experimentation in order to assess hypotheses on large populations, and facilitate the development of ex-perimental protocols to investigate the influence of socio-cultural background on soundscape perception.