The noise pollution in cities is a continuously developing problem [1]. The effect of street dimensions, traffic density and composition on the noise level in urban canyons has been investigated. A series of noise measurements, were made in "canyon" streets in Thessaloniki. The main purpose of the measurements was to examine the correlation between urban road traffic noise, traffic volume, traffic composition and source - receiver distance. Urban road traffic noise monitoring has been carried out in thirty-seven different sites, representative of different urban configurations, where sound levels were measured and at the same time, the number and the type of the passing vehicles were taken. Measurements have been performed for fifteen minutes, taking into account the vacancy for the correlated hourly L-Aeq,L-h levels. Part of research constituted the relation of each category of vehicles; light, heavy and motorised two-wheeled vehicle, with the level of urban road traffic noise. We created the optimally connected mathematical expression which determines the degree of influence of each vehicle category to attribute the best cross-correlation with the measured urban road traffic noise levels. They were developed four models using linear regression analysis taking into account the most important road traffic noise factors: traffic volume, traffic composition and the source - receiver distance. The traffic composition had been classified by three categories. The survey has showed that the noise levels in urban traffic noise evaluation, e.g. the hourly equivalent noise level L-Aeq,L-h, can be estimated by knowing the traffic composition with reasonably better precision than by making use of the traffic flow. It is also given the possibility to compare the results of the regression analysis for developed models. Also it was mathematically approved that the number of motorised two - wheeled vehicles influences substantially the configuration of the final noise level, but not as much as the heavy vehicles.
