Dynamic noise maps are computed to represent the noise levels generated by road traffic in realtime to determine the population exposure to environmental noise. To this aim, a network ofacoustic sensors can be deployed in representative street locations to capture and process rawacoustic data before estimating the equivalent noise levels (Leq) of the maps. In the frameworkof the DYNAMAP project, we conducted a study of the acoustic levels of several streets to de-termine the most suitable places for the sensors to be deployed in the case study of Milan city.The conclusions of that analysis were that the streets under test could be described by two dif-ferent time-dependent behaviors, instead of the functional classification of roads. For a properevaluation of the equivalent valueLeq, several acoustic events existent in the street measurementshave to be removed, since they are not generated by road traffic noise. These events, denoted asanomalous noise events, can come from sirens, horns, noisy human activities, etc. To that effect,the measured acoustic signal in the street has to be studied to classify whether the sample belongsto regular road traffic noise, or it is an anomalous noise event, assuming that the typology of theseevents is very diverse. In order to allow the deployment of the monitoring system, this evaluationshould be conducted previously, and if possible, in the acoustic sensor sites resulting from theprevious analysis. In this paper, we evaluate the different typologies of anomalous noise eventsobserved in the city of Milan and we describe them statistically according to the identified streetclusters. Moreover, we study the potential impact they may cause in theLeqevaluation if notdiscarded.